Falcon deep research failed twice (falcon 600 s timeout with unavailable perplexity-lite fallback; then exit 137 at --timeout 2400) and succeeded on the third attempt (IFT172-deep-research-falcon.md). Additions (primary papers not cached): endogenous IFT172 at ciliary base and along the axoneme in human fibroblasts; purified IFT172 binds and remodels phospholipid membranes (Wang 2018), with membrane binding and IFT57 binding mutually exclusive [file:human/IFT172/IFT172-deep-research-falcon.md "Membrane association and binding of IFT57 to IFT172 appear mutually exclusive, suggesting a possible way to regulate when IFT172 associates with membranes versus an assembling IFT complex."]; Chlamydomonas ift172 ts mutant accumulates IFT material at the tip (turnaround role); a C-terminal U-box-like ubiquitin-binding domain (Zacharia, eLife) with no established E3 activity. None is asserted as a GO MF here; membrane binding raised as a suggested question.
IFT172 is not in HPA v25 (no cilium/centrosome call; "not in HPA"), so the atlas neither supports nor contradicts its module placement; there are no GO_REF:0000052 rows. The module's assignment (IFT-B peripheral subunit of anterograde trains; anterograde transport and cilium assembly) rests entirely on biochemical, model-organism and human-genetic literature, which supports it. Core functions agree; no cargo-adaptor MF is asserted for IFT172.
Every IFT-B subunit annotated from PMID:26980730 (Katoh et al. 2016, an IFT-B architecture paper) carries an IDA to GO:0030991 intraciliary transport particle A. This contradicts the paper's own content (which supports the GO:0030992 IFT-B IPI rows from the same paper) and the established biochemistry. Marked REMOVE as a genuinely contradicted claim (likely batch term-selection error), not as second-guessing assay data.