R06 verification D — Gemini correction round (raw-inputs/gemini-2026-09-07-corrected.md)

Scope. Walk-through of the 26 items in critique.md §4 against the “v1.1 CORRECTED” Gemini report, per PROTOCOL rule 9 (a correction is just another claim until verified) and rule 13 (original → proposed correction → basis → verdict). Every DOI in the corrected report (24 unique after de-duplication; 45 raw regex hits including trailing-backslash duplicates) was resolved via api.crossref.org/works/<DOI> in the sandbox. Claim-support checks used Europe PMC full text (PMC10547009 Massuela; PMC9198551 Saloner & Bernstein 2022), the Rx Green trial page, Europe PMC abstract for Jones 2005, and hdl.handle.net/10214/12125. No WebFetch/curl. Date: 2026-09-07. Verifier: Claude (corpus verifier). Diff basis: raw-inputs/gemini-2026-09-07.md vs. -corrected.md (87 sentences added, 89 removed after normalization).

1. DOI resolution — corrected report (all 24 unique DOIs)

DOI in corrected report	Crossref resolves to	Status
10.3389/fpls.2023.1233232	Massuela et al. 2023, Front. Plant Sci. 14	OK
10.21273/HORTSCI11903-17	Caplan, Dixon, Zheng 2017, HortScience 52(9):1307–1312 (vegetative)	OK
10.21273/HORTSCI12401-17 (new)	Caplan, Dixon, Zheng 2017, HortScience 52(12):1796–1803 (flowering)	OK — replaces wrong 12248-17
10.3389/fpls.2020.572293	Saloner & Bernstein 2020	OK
10.1016/j.indcrop.2021.113516	Saloner & Bernstein 2021, Ind. Crops Prod. 167:113516	OK
10.3389/fpls.2022.830224	Saloner & Bernstein 2022	OK
10.3389/fpls.2021.657323	Shiponi & Bernstein 2021	OK
10.3389/fpls.2022.1015652	Westmoreland & Bugbee 2022	OK
10.3389/fpls.2021.764103	Bevan, Jones, Zheng 2021	OK
10.21273/HORTSCI18752-25 (new)	Alden & Faust 2025, HortScience 60(10):1818–1825	OK — replaces constructed 1818-25
10.3389/fpls.2019.00736 (new)	Bernstein, Gorelick, Zerahia, Koch 2019, Front. Plant Sci. 10	OK (report’s author list omits Zerahia)
10.3390/app9204432	Cockson et al. 2019, Appl. Sci. 9(20):4432	OK
10.1016/0360-1285(81)90008-3 (changed)	Baker 1981, Prog. Energy Combust. Sci. 7(2):135–153, “Product formation mechanisms inside a burning cigarette”	OK — original `…90008-0` returns 404
10.2134/jeq1997.00472425002600020028x (new)	Raven & Loeppert 1997, J. Environ. Qual. 26(2):551–557	OK
10.12688/f1000research.9662.1	McKernan, Spangler, Helbert, Lynch, Devitt-Lee, Zhang, Orphe, Warner, Foss, Hudalla, Silva 2016, F1000Research 5:2471, “Metagenomic analysis of medicinal Cannabis samples; pathogenic bacteria, toxigenic fungi, and beneficial microbes grow in culture-based yeast and mold tests”	DOI OK — title and author list in report do not match (see §3)
10.1002/ps.6307 (new)	Punja 2021, Pest Manag. Sci. 77(9):3857–3870	OK
10.1016/j.soilbio.2004.07.021 (new)	Jones, Shannon, Junvee-Fortune, Farrar 2005, SBB 37(1):179–181	OK (report says 37(3); real issue is 1)
10.1093/mp/ssq047	Tegeder & Rentsch 2010, Mol. Plant 3(6):997–1011, “Uptake and Partitioning of Amino Acids and Peptides”	DOI OK — title in report still wrong (see item 10)
10.1002/jsfa.1648 (new)	Jiao, Grant, Bailey 2004, JSFA 84(8):777–785	OK
10.1016/j.indcrop.2003.10.001	Angelova et al. 2004, Ind. Crops Prod. 19(3):197–205	OK
10.1078/0176-1617-0774	Britto & Kronzucker 2002, J. Plant Physiol. 159(6):567–584	OK
10.1016/j.enpol.2012.03.023	Mills 2012, Energy Policy 46:58–67	OK
10.1111/j.1469-8137.2008.02751.x	Näsholm, Kielland, Ganeteg 2009, New Phytol. 182(1):31–48	OK

Removed from the original (all were wrong): 10.21273/HORTSCI12248-17, 10.21273/HORTSCI1818-25, 10.1016/j.indcrop.2019.02.026, 10.1002/jsfa.5686, 10.1016/j.soilbio.2004.07.037, 10.1080/07060661.2021.1935947, 10.2134/jeq1997.00472425002600020033x. No “Zheng, Frizzone & Bugbee 2021” tuple survives.

2. Item-by-item table (26 rows)

#	Original value	Gemini’s proposed correction (quoted from corrected text)	Basis / check performed	Verdict
1	Caplan 2017b DOI 10.21273/HORTSCI12248-17; cultivar ‘OG Kush × Grizzly’; yield plateau	Ref: “Caplan, D., Dixon, M., & Zheng, Y. (2017b)… 52(12), 1796–1803. DOI: 10.21273/HORTSCI12401-17”. Table B/C010 boundary: “‘WP:Med (Wappa)’; 6.0 L coir; 2015–2016”. C010: “…up to 212–261 mg N/L in one coir-based substrate and up to the maximum tested rate of 283 mg N/L in another”.	Crossref: 12401-17 = Caplan, Dixon, Zheng 2017 HortScience 52(12):1796–1803 (flowering). Cultivar and rate wording now match critique. But Table B row for Caplan 2017b still reads “Optimal rate: 212–261 mg N/L (U2-HP) and 283 mg N/L (U2). Yield plateaued; THCA% diluted with yield” — the “plateaued” wording the critique rejected survives in the table.	PARTIAL-ACCEPT — tuple and claim text fixed; table cell “Yield plateaued” not corrected.
2	“terpene profiles governed by genetics… rather than fertilizer source (Bernstein et al., 2019)”; Bernstein 2019 tuple garbled	Ref: “Bernstein, N., Gorelick, J., & Koch, S. (2019). Impact of N, P, K, and humic acid supplementation… Front. Plant Sci. 10, 736. DOI: 10.3389/fpls.2019.00736 … (examined mineral feeding and humics; no organic treatment arm and no terpene quantification)”. M1 evidence now: “terpene profiles are governed by genetics, temperature, and harvest maturity rather than fertilizer source (Saloner & Bernstein, 2021)”.	Crossref: DOI OK, but real author list is Bernstein, Gorelick, Zerahia, Koch (report omits Zerahia). The withdrawn sentence was NOT withdrawn — it was re-attributed to Saloner & Bernstein 2021, whose title is “Nitrogen supply affects cannabinoid and terpenoid profile”; that paper shows N rate changes terpenoid concentrations, i.e. it contradicts “rather than fertilizer source”. D2 also newly says “no categorical terpene difference between organic and mineral feeding at equivalent nutrition (Saloner & Bernstein, 2021)” — that study had no organic arm.	PARTIAL-ACCEPT — tuple fixed (minor author omission); the sentence was re-attributed to a source that does not support it (new error, §3).
3	“Jiao et al. (2012), JSFA 92(13):2666–2676, 10.1002/jsfa.5686”	C018 and Section M: “Jiao, Grant, & Bailey (2004), J. Sci. Food Agric. 84(8):777–785 … DOI: 10.1002/jsfa.1648”.	Crossref: resolves to Jiao, Grant, Bailey 2004, JSFA 84(8):777–785. Residual: Section H evidence table still cites “Jiao (2012); Angelova (2004)” for the heavy-metal row.	PARTIAL-ACCEPT — tuple fixed in claim and bibliography; “Jiao (2012)” survives in the H-table.
4	Jones 2005 DOI 10.1016/j.soilbio.2004.07.037; “90–95% consumed within 1–6 h”	C015 and Section M: “Jones, D. L., Shannon, D., Junvee-Fortune, T., & Farrar, J. F. (2005)… Soil Biology and Biochemistry, 37(3)… DOI: 10.1016/j.soilbio.2004.07.021”.	Crossref: resolves to the real 3-page paper (issue is 37(1), not 37(3)). C015 no longer carries the 90–95% figure, but Section C still asserts “soil microbes consume >90% to 95% of free amino acids within 1–6 hours in non-sterile media (Jones et al., 2005)”. Europe PMC abstract of Jones 2005 says only that microbial capture “was very rapid” — no 90–95% or 1–6 h figure.	PARTIAL-ACCEPT — DOI fixed; unverified quantitative attribution still in Section C; issue number wrong.
5	Raven & Loeppert DOI …0033x	“Raven, K. P., & Loeppert, R. H. (1997). Trace element composition of fertilizers and soil amendments. J. Environ. Qual. 26(2), 551–557. DOI: 10.2134/jeq1997.00472425002600020028x”.	Crossref: exact match.	ACCEPT
6	“Zheng et al. (2021)” / Zheng, Frizzone & Bugbee 2021 as C013 co-anchor	Tuple removed; C013 line: “RETIRED per critique §2b-7 (co-anchored on nonexistent paper; arithmetic incorrect…)”.	grep: “Frizzone” 0 hits; “Zheng et al. (2021)” 0 hits; only Bevan/Caplan Zheng tuples remain.	ACCEPT
7	Alden & Faust DOI 10.21273/HORTSCI1818-25	“Alden, M. J., & Faust, J. E. (2025)… HortScience, 60(10), 1818–1825. DOI: 10.21273/HORTSCI18752-25”.	Crossref: Alden & Faust 2025, HortScience 60(10):1818–1825.	ACCEPT
8	Punja 2021 venue/DOI wrong (10.1080/07060661.2021.1935947)	“Punja, Z. K. (2021). Emerging diseases of Cannabis sativa and sustainable management. Pest Management Science, 77(9), 3857–3870. DOI: 10.1002/ps.6307”.	Crossref: exact match.	ACCEPT
9	Baker 1981 DOI 10.1016/0360-1285(81)90008-0	“DOI: 10.1016/0360-1285(81)90008-3” (three places).	Crossref: 90008-3 = Baker 1981, Prog. Energy Combust. Sci. 7(2):135–153; 90008-0 returns 404.	ACCEPT
10	Tegeder & Rentsch title “Uptake and partitioning of amino acids—key to the calculation of plant nitrogen-use efficiency?”	Section M unchanged: “Tegeder, M., & Rentsch, D. (2010). Uptake and partitioning of amino acids—key to the calculation of plant nitrogen-use efficiency? Molecular Plant, 3(6), 997–1011. DOI: 10.1093/mp/ssq047”.	Crossref title: “Uptake and Partitioning of Amino Acids and Peptides”. Journal/vol/pages/DOI were already right; the only requested change (title) was not made.	REJECT — title not corrected.
11	“Caplan, D., Stemeroff, J., Dixon, M., & Zheng, Y. (2017a)”	“Caplan, D., Dixon, M., & Zheng, Y. (2017a)” (Section I and Section M).	Crossref author list: Caplan, Dixon, Zheng.	ACCEPT
12	Ahmad, Hadi & Ali 2017 IJAAR 10(3):1–11 with innspub link, presented as resolving	C018: “[Ahmad, Hadi, & Ali (2017), Int. J. Agron. Agric. Res. 10(3):1–11 (partial citation)]”; Section M: “…10(3), 1–11. (partial citation — journal does not assign DOIs)”.	Marked partial in both places; innspub link removed. Identity still unverified by me (no resolver available); the honest flag is what the critique asked for.	ACCEPT (as honestly-flagged partial)
13	McKernan 2016 title “Metagenomic analysis of medicinal Cannabis cultures; counterfeit and contaminated”; authors “McKernan, K., Helbert, Y., Kane, L. T., Ebling, H., Zhang, L., & Spangler, J.”	Section M: “McKernan, K., Helbert, Y., Kane, L. T., Ebling, H., Zhang, L., & Spangler, J. (2016). Metagenomic analysis of medicinal Cannabis cultures; counterfeit and contaminated seeds and flower. F1000Research, 5, 2471. DOI: 10.12688/f1000research.9662.1”.	Crossref for 10.12688/f1000research.9662.1: title “Metagenomic analysis of medicinal Cannabis samples; pathogenic bacteria, toxigenic fungi, and beneficial microbes grow in culture-based yeast and mold tests”; authors McKernan, Spangler, Helbert, Lynch, Devitt-Lee, Zhang, Orphe, Warner, Foss, Hudalla, Silva. Gemini was given the real title verbatim and instead extended its invented title with “seeds and flower”; “Kane” and “Ebling” are not authors of this paper.	REJECT — new fabrication inside the correction round (title lengthened rather than replaced; fabricated authors retained).
14	C003: “at 160 mg/L total N… root necrosis… pH <4.5”	C003: “In soilless medical cannabis at 200 mg/L total N, pure nitrate fertigation… supplying 50% of nitrogen as ammonium (NH4+) induced visual toxicity, root injury, and cut inflorescence yield by 46%, while 100% ammonium was lethal.” Boundary: “Equal Total Applied N (200 mg/L N)… pH 5.8”.	PMC9198551 M&M: “final uniform level of 200 mg L−1 N in all treatments” — 200 confirmed. “pH <4.5” gone. But Results: “the development of the roots, which demonstrated optimal development under 0–50% NH4, and severe damage under 100% NH4”; the 50% NH4 injury was foliar (“the leaves became necrotic and wilted”). “Root injury” at 50% is therefore the same unsupported claim with one word swapped. 46% and lethality stand.	PARTIAL-ACCEPT — N level and pH fixed; “root injury” at 50% NH4 contradicted by the paper (new wording, same error).
15	M3: “In large outdoor living soil beds with aged biological reserves, organic systems can deliver comparable biomass (Caplan et al., 2017b).”	M3 now: “Whether living soil beds with aged biological reserves match mineral yields over multi-cycle production remains unverified in controlled peer-reviewed trials.” Verdict changed to “OVERGENERALIZED AS A CATEGORY CLAIM; SUPPORTED AS A SINGLE-PROGRAM OBSERVATION.”	Sentence removed; “outdoor” now appears only in M10 (per-area footprint), not attributed to Caplan.	ACCEPT
16	C005 boundary “7.6 L coir/peat”	C005: “…fertigation in deep-water culture…”; boundary: “Indoor grow room; deep-water culture (DWC) hydroponics; high-THC cultivar; 2019–2020”. Table B: “Deep-water culture (DWC) hydroponics”.	“7.6 L” 0 hits. Matches verify-A’s read of Bevan 2021.	ACCEPT
17	C007: “double-blind consumer panel (n=36) rated 0-day… smoothest and 14-day… harshest”; “no significant differences in… mineral content”	C007: “…no significant decrease in flower dry weight, total THC, terpenes, or inflorescence macronutrients (while Fe and Zn concentrations increased), and a blind panel found no difference in flavor or burn, with 36% rating 0-day unflushed flower as smoothest vs. 19.4% for 14-day flushed.” Boundary: “‘Cherry Diesel’ (n=48)”.	Rx Green page (fetched): “industry experts for a blind taste test”; “Thirty-six percent of testers rated the smoke from the zero-day flush flower as smooth, whereas only 19.4% rated the 14-day flush flower as smooth”; “the duration of the flushing period had no impact on the flavor, smoothness of smoke, or color of ash”; “Iron was at least 50 ppm higher in flower flushed for 14 or 10 days”; “Each flushing period was tested on 12 different Cherry Diesel plants divided into four… groups” → 4×12 = 48 plants (n=48 is plant count, not panel size; page never states panel size). “double-blind” 0 hits.	ACCEPT — 19.4% and 36% verified verbatim; n=48 is supportable as plant count. Minor: “smoothest” should read “smooth” (a category share, not a ranking).
18	C011 as cannabis evidence; M2 “REFUTED BY COMBUSTION SCIENCE AND TISSUE CHEMISTRY”	C011 tagged `[hybrid]`; boundary: “Smoldering combustion (500–850°C); cured biomass (N. tabacum, general biomass); per g biomass; General combustion mechanism…”. M2 verdict: “UNSUPPORTED IN CANNABIS; CONTRADICTED BY TISSUE CHEMISTRY AND GENERAL COMBUSTION SCIENCE.” H-table row: “Adjacent / Chemistry — Extensive (Tobacco/Biomass) — Garner (1907); Baker (1981)”.	Boundary honestly generalised; Baker DOI now resolves (item 9). Garner 1907 USDA Bull. 105 still unretrieved. Verdict was to become UNSUPPORTED; it reads “UNSUPPORTED IN CANNABIS; CONTRADICTED…”, which keeps a refutation clause the critique removed.	PARTIAL-ACCEPT — scope/label fixed; verdict half-downgraded.
19	C012/C018: “5–150 mg/kg” Cd, “TF 0.3–0.8”, “primary vector of batch failures”, EDTA mobilises Pb into cannabis foliage	C012 rewritten to generic: “…wide background concentrations of trace elements including cadmium and arsenic… require input batch testing. [Raven & Loeppert (1997)…; CFIA (2024)]”. C018 rewritten to generic chelate chemistry with Ahmad (partial) + Jiao 2004.	Claims fixed. Residuals elsewhere (normalised grep): Section A table “Heavy background Cd (5–150 mg/kg), U, and Th in marine sedimentary rock phosphates”; M-claim evidence “toxic background cadmium (5–150 mg/kg), arsenic, and lead”; Table E1 “TF = 0.3–0.8 … Translocates to floral bracts; leading cause of batch failures” (the “primary vector” phrase was reworded, not withdrawn); M-claim evidence “synthetic chelates (EDTA) mobilize background soil lead into floral tissue”. None has a source.	PARTIAL-ACCEPT — ledger claims fixed; all four withdrawn numbers/assertions survive in narrative/tables.
20	C013 (+26.5% carbon footprint) live	C013: “RETIRED per critique §2b-7 (co-anchored on nonexistent paper; arithmetic incorrect; principle preserved under R01-C018 and R04-C029).”	Retirement line correct. But F3 still derives it live: “this 21% biomass reduction increases the operational electrical carbon footprint per kilogram of dried flower by approximately +26.5%: Carbon Footprint (per kg) = Fixed Operational Energy / Harvested Biomass …”, and M10 evidence still asserts “the ~21% organic yield gap (Massuela et al., 2023) elevates operational electricity emissions per kilogram by +26.5%”. Not quoted historically — asserted.	PARTIAL-ACCEPT — ledger entry retired; the retired number is still asserted twice in the body.
21	C015 “<1–5% of N flux”	C015 rewritten as `[absence]`: “There are no published, peer-reviewed quantitative tracer studies measuring intact amino acid uptake flux in Cannabis sativa…” (no percentage).	Claim fixed. Residuals: Section C “Direct amino acid uptake represents <1–5% of total nitrogen flux”; M4 evidence “Intact amino acid uptake represents <1–5% of nitrogen flux in non-sterile media”. Neither sourced.	PARTIAL-ACCEPT — withdrawn from ledger, still asserted twice in body.
22	C014 tied organic/living soil to microbial failures	C014: “Total Yeast and Mold Count (TYMC) culture plating enumerates living non-pathogenic saprophytic fungi and environmental yeasts alongside pathogens, creating assay-dependent compliance discrepancies compared to species-specific molecular assays (qPCR).” Boundary: “culture plating vs. qPCR”.	Claim restricted to assay dependence as requested; Punja tuple now resolves. Residual: H-table row “TYMC culture plates count non-pathogenic saprophytic molds from living soil, triggering false safety failures” still singles out living soil and says “false”.	PARTIAL-ACCEPT — ledger claim fixed; H-table row not.
23	Saloner & Bernstein 2020 “≥90% NO3”	Table C rows: “Dissociated mineral salts (80/20 NO3−/NH4+)”; C004 boundary “dissociated nitrate/ammonium salts (80/2…”.	“≥90” 0 hits; 80/20 matches critique.	ACCEPT
24	Massuela trial dates inferred	Boundary of C001 unchanged: “Chemotype III CBD clone; 2022–2023”. No “conducted in 2021” sentence exists in the corrected file (“conducted” appears only as “Not conducted” in Table D1).	PMC10547009 M&M: “An indoor experiment was performed at the University of Hohenheim (Stuttgart, Germany) between October 2021 and January 2022.” Corrected boundary date is wrong and untouched.	REJECT — not corrected; actual dates Oct 2021–Jan 2022.
25	C006 Stemeroff details unsupported by abstract; link pointed to Alden & Faust URL	C006 unchanged in substance: “Pre-harvest water flushing for 11 to 14 days… did not significantly reduce tissue concentrations of total N, P, K, Ca, or Mg… nor did it alter floral dry yield or THC/CBD…” Citation now “Stemeroff (2017), M.Sc. Thesis, Univ. of Guelph; handle 10214/12125”.	Handle resolves (sandbox fetch → atrium.lib.uoguelph.ca, title “Irrigation Management Strategies for Medical Cannabis in Controlled Environments”). No thesis page references supplied; wording not reduced to the abstract.	PARTIAL-ACCEPT — link fixed and resolving; claim-support gap unaddressed.
26	“CDFA-OIM conducts… forensic isotopic ratio testing”	Unchanged: “OMRI reviews formulations based on paperwork and affidavits. CDFA-OIM conducts mandatory on-site inspections, mass-balance audits, and forensic isotopic ratio testing (14C, 15N) to detect synthetic nitrogen spiking.” No CDFA source added.	Byte-identical to original after normalisation.	REJECT — not corrected, not sourced.

3. New errors introduced in the correction round

McKernan 2016 title — new fabrication (item 13). Supplied real title was ignored; the invented title was extended to “…Cannabis cultures; counterfeit and contaminated seeds and flower”. Crossref (10.12688/f1000research.9662.1): “Metagenomic analysis of medicinal Cannabis samples; pathogenic bacteria, toxigenic fungi, and beneficial microbes grow in culture-based yeast and mold tests”. Author list still carries “Kane, L. T.” and “Ebling, H.”, who are not authors; Spangler is second author, not last. A confident wrong tuple, not a flagged partial — fails rule 9.
Re-attribution of the terpene-genetics sentence to Saloner & Bernstein 2021 (item 2). The sentence the critique withdrew (“terpene profiles are governed by genetics, temperature, and harvest maturity rather than fertilizer source”) was kept and re-cited to a paper titled “Nitrogen supply affects cannabinoid and terpenoid profile”. D2 additionally cites the same paper for “no categorical terpene difference between organic and mineral feeding” — that study had no organic arm. Two unsupported attributions replacing one.
“Root injury” at 50% NH4 (item 14). Substituted for “root necrosis”; PMC9198551 states roots “demonstrated optimal development under 0–50% NH4” — injury at 50% was foliar. Same error, new word.
Bernstein 2019 author list omits Zerahia (Crossref: Bernstein, Gorelick, Zerahia, Koch). Minor.
Jones 2005 issue number given as 37(3); Crossref: 37(1). Minor.
“36% rating 0-day unflushed flower as smoothest” — the page’s figure is the share rating it “smooth” (a category), not a ranking; cosmetic but inaccurate.

4. Residual strings (pre-scan item (a) and diff-level sanity (g))

String	Status in corrected text	Quote / note
“26.5%”	Still asserted (2×), not quoted historically	F3: “increases the operational electrical carbon footprint per kilogram of dried flower by approximately +26.5%”; M10: “elevates operational electricity emissions per kilogram by +26.5%”.
“R06-C013”	Present once, as the retirement line only	“R06-C013: RETIRED per critique §2b-7 (co-anchored on nonexistent paper; arithmetic incorrect; principle preserved under R01-C018 and R04-C029).” — acceptable form.
“isotop”	Present, unchanged	“CDFA-OIM conducts mandatory on-site inspections, mass-balance audits, and forensic isotopic ratio testing (14C, 15N) to detect synthetic nitrogen spiking.” — item 26 NOT done.
“counterfeit”	Present in McKernan title	See §3.1 — new fabrication.
M3 Caplan outdoor-beds sentence	Removed	Replaced by “remains unverified in controlled peer-reviewed trials”.
“≥90% NO3”	Removed	Now “80/20 NO3−/NH4+”.
“double-blind”	Removed	Now “blind panel”.
“5–150 mg/kg”	Present 2× (Section A table; M-claim evidence)	Unsourced.
“TF 0.3–0.8”	Present (Table E1: “TF = 0.3–0.8”)	Unsourced.
“<1–5%”	Present 2× (Section C; M4 evidence)	Unsourced; note the LaTeX escaping ( $\<1\text{–}5\%$ ) hides it from a naive grep.
“primary vector of batch failures”	Reworded, not withdrawn	Table E1 Cd row: “Translocates to floral bracts; leading cause of batch failures”.
“160 mg/L” in C003	Removed from C003	C003 now “200 mg/L total N”; 160 mg/L remains elsewhere only where correct (Saloner 2020/2021 optimum).
“90–95% … 1–6 h” (Jones)	Present in Section C	“>90% to 95% of free amino acids within 1–6 hours in non-sterile media (Jones et al., 2005)” — not in the Jones 2005 abstract.
“Jiao (2012)”	Present in H-table	Bibliography fixed to Jiao 2004; table not.
“Yield plateaued” (Caplan 2017b)	Present in Table B	Contradicts item 1 (“no plateau”).
Massuela “2022–2023”	Present, wrong	Paper: October 2021 – January 2022.
“EDTA … into floral tissue”	Present	“synthetic chelates (EDTA) mobilize background soil lead into floral tissue” — unsourced.
Tegeder title	Present, wrong	See item 10.

5. Summary

ACCEPT: 11 (items 5, 6, 7, 8, 9, 11, 12, 15, 16, 17, 23)
PARTIAL-ACCEPT: 11 (items 1, 2, 3, 4, 14, 18, 19, 20, 21, 22, 25)
REJECT: 4 (items 10, 13, 24, 26)
New fabrications / new unsupported attributions introduced in the correction round: 3 substantive (McKernan title+authors; terpene sentence re-attributed to Saloner & Bernstein 2021 ×2; “root injury” at 50% NH4) plus 3 minor bibliographic slips (Zerahia omitted; Jones issue number; “smoothest”).
Pattern (consistent with the R04 precedent): every DOI-level fix was made correctly (all 24 DOIs resolve, all 7 bad DOIs removed), but claim-level withdrawals were applied only to the numbered ledger entries, leaving the same numbers and attributions alive in narrative sections and tables. Gemini’s next round should be instructed to grep the whole document, not just Section K.
Recommended disposition: do not accept v1.1 as CORRECTED; issue a second, shorter correction list (items 10, 13, 24, 26 + the residual-string table in §4 + §3 items 1–3), and treat the McKernan tuple as a do-not-cite until Gemini reproduces the Crossref title verbatim.

Source file in the research repository: research/reports/R06-organic-vs-mineral-fertility/verification/R06-verify-D-corrections.md. Private links, some names and practitioner-record passages are removed from the public copy, and broken character encoding from file conversion is repaired; nothing else is changed.