table-mismatch 4 · prose-contradiction 0 · confidence-mismatch 4 · coverage-mismatch 1. Report-only.
Systemic pattern — TDS “in_progress” boilerplate, continued: The following pages carry “Contamination Profile State: in_progress for analytes measured in the TDS file” even though those analytes are status: populated with correct values in frontmatter and body table: frankfurter, frozen-peas, ground-beef, honey, iceberg-lettuce, ketchup, lamb-chop. Not individually rowed; body tables correct; mismatch confined to status-section boilerplate. Same systemic pattern noted in sessions 2 and 3.
New observation — fruit-juice blank-body-table (P1): fruit-juice follows the same blank-table pattern as black-pepper (session 1), bread, and cashews (session 2): body table all ”—” while frontmatter has populated values. Fruit-juice has sources:26 (rich page), confirming this is a generator-timing issue rather than a stub-class problem.
New observation — fromage-frais/custard identical contamination_profile values: fromage-frais has identical frontmatter values to custard (Pb [0,14.8] p95=51.1, Cd [0,4.3], iAs [0,11.2], tHg [0,1], Al [0,1734], Sn [0,200.6]) with sources:0 for both pages. Both are FSA-sourced stubs. Flagged as a potential copy-paste error during FSA stub generation. Reported in session 3 for custard; carried forward as context for fromage-frais finding below.
| Page | Analyte | Frontmatter value | Prose / table claim | Mismatch type | Which looks stale |
|---|---|---|---|---|---|
| flour-tortilla | all populated analytes | n_studies: 2 per analyte (Pb, Cd, tAs, tHg, Ni, Cr, U); sources: 1 (sole source: fda2022-tds-elements-fy2018-fy2020) | Body table shows n=2 for all populated analytes; only 1 source in legend; methodology note states TDS counts as 1 source | P1 table≠frontmatter | n_studies in frontmatter — same TDS sources:1/n_studies:2 pattern as bran-cereal-with-raisins, chicken-noodle-soup, chicken-potpie, collard-greens, cream-of-wheat (prior sessions) |
| fruit-juice | Pb, tAs | Pb: status:populated; typical [0.5,5] p95_ppb:null; n_studies:2; confidence:medium. tAs: status:populated; typical [1,10] p95=null; n_studies:2; confidence:medium. sources:26 (rich page) | Body table shows ”—” for every analyte (all 10 analytes render as blank/dash including the two populated ones) | P1 table≠frontmatter | Body table — generator ran before or without Pb/tAs being populated; same blank-table pattern as black-pepper (session 1), bread and cashews (session 2) |
| iceberg-lettuce | all populated analytes | n_studies: 2 per analyte (Pb, Cd, tAs, tHg, Ni, Cr, U); sources: 2; body table cites only fda2022-tds-elements-fy2018-fy2020 for all populated analytes | Body table shows n=2 for all analytes; only 1 data source cited; methodology note states TDS counts as 1 source | P1 table≠frontmatter | n_studies in frontmatter — same TDS sources:1-data-source/n_studies:2 pattern; sources:2 in frontmatter but second source contributes no data rows |
| lamb-chop | all populated analytes | n_studies: 2 per analyte (Pb, Cd, tAs, tHg, Ni, Cr, U); sources: 1 (sole source: fda2022-tds-elements-fy2018-fy2020) | Body table shows n=2 for all populated analytes; only 1 source in legend | P1 table≠frontmatter | n_studies in frontmatter — clearest single-source instance of the TDS sources:1/n_studies:2 pattern |
| freshwater-fish | Pb, Cd, iAs, tAs, tHg, Ni, Al, Cr (8 analytes) | Pb n=22; Cd n=19; iAs n=6; tHg n=33; Ni n=10; Al n=5; Cr n=8; tAs n=16 — all status:populated with real multi-study values | ”Contamination Profile State: All ten contamination_profile sub-blocks are pending” | P2 confidence-mismatch | Prose — profile state section not updated after synthesis; 8 analytes are populated with 5–33 studies each |
| fromage-frais | Pb, Cd, iAs, tHg, Al, Sn | Pb [0,14.8] p95=51.1 n=1; Cd [0,4.3] p95=5.2 n=1; iAs [0,11.2] p95=33.6 n=1; tHg [0,1] p95=2.4 n=1; Al [0,1734] p95=2352 n=1; Sn [0,200.6] p95=245.4 n=1; sources:0 | Intro: “Exact concentration values remain in progress until Table 6 is parsed”; “Contamination Profile State: in_progress” | P2 confidence-mismatch | Prose — body table correctly shows values attributed to fsa2016-infant-food-formula-metals-survey; intro and profile state not updated after synthesis; note sources:0 while values exist and are identical to custard — potential copy-paste error during FSA stub generation |
| infant-formula-powder | Pb, Cd, iAs, tAs | Pb [0.5,15] p95=50 n=3; Cd [0.1,2] p95=3.5 n=3; iAs [0.5,3] n=3; tAs [2,12] p95=13 n=4 — all status:populated | Intro: “This is a structural ingredient node created so product pages can link to a real wiki target. Occurrence values remain pending until a source is promoted for this ingredient.”; profile state: “The machine-readable contamination profile is pending” | P2 confidence-mismatch | Prose — 4 analytes populated from 3–4 sources each; intro and profile state not updated after synthesis |
| leafy-greens | Pb, Cd, tAs, tHg, Ni, Al, Cr, Sn (8 analytes) | Pb n=19 [5,200] p95=610 high; Cd n=22 [5,150] p95=400 high; tAs n=11 [5,250] p95=410 high; Ni n=13 [35,560] p95=3465 high; Cr n=15 [10,560] p95=1551 high; Al n=3 [1000,50000] p95=150000 medium; tHg n=6 [1,15] p95=50 medium; Sn n=2 [0,10] low | Intro: “This is a structural ingredient node created so product pages can link to a real wiki target. Occurrence values remain pending until a source is promoted for this ingredient.”; profile state: “pending” | P2 confidence-mismatch | Prose — 8 analytes populated from sources:38 with 2–22 studies each; intro and profile state not updated after synthesis; body table correctly reflects frontmatter |
| ketchup | Sn | status:populated; typical_ppb:[null,null]; n_studies:0; body table shows “data gap” | Prose (“Why this commodity accumulates heavy metals”): “The second and historically more significant pathway is tin migration from can interiors… The EU limits Sn in canned solid foods at 200 mg/kg” — Sn described as a major contamination mechanism for this acidic canned product | P3 coverage-mismatch | Frontmatter — prose gives Sn a prominent mechanistic role; no occurrence data is in the corpus (data gap is correct), but the substantive prose discussion creates an expectation that the analyte is characterized |