Fixed shape fits give different alternate event depths for three TESS candidates in sectors 94 and 104

We tested whether short period TESS candidates have different depths in alternate events. We screened 786 NASA TOI entries with periods of 0.15–2.0 days and PC or APC status on 2026-10-21. The screen used the latest sector in frozen MAST responses and gave 630 finite descriptive ratios. We selected three cases for detailed tests. We used fixed catalog periods and an exposure averaged shape. The even-minus-odd depth difference for TOI 3338.01 is +3586 ± 671 ppm in sector 104. TOI 6782.01 gives −7286 ± 1256 ppm in sector 94. TOI 6805.01 gives −2939 ± 722 ppm in sector 94. Errors measure scatter between fitted events under an independent event assumption. SAP fits and 12 single parameter changes preserve each sign. Older QLP sectors do not establish a stable pattern. A shared shape with a quadratic baseline weakens TOI 6782.01 to an absolute difference divided by error of 1.52. TOI 6805.01 reaches the allowed V shape limit. Four ground files for TOI 3338.01 cannot isolate parity from instrument and filter effects. TOI 6782.01 has a prior unvetted TESS Ten Thousand Catalog entry. A QLP sector 104 file with a source timing warning is excluded from scientific comparisons; the main result uses SPOC. The results identify follow-up leads. They do not confirm planets, binaries, or doubled orbital periods. The package preserves code, twelve FITS files, frozen selection data, results, source hashes, and reproduction instructions. The errors exclude model uncertainty and survey selection effects.

TESS · TOI 3338.01 · TOI 6782.01 · TOI 6805.01 · alternate event depths · eclipsing binaries · candidate vetting · public astronomical data · reproducible research