Consistency: CONTRADICTION on GO:0000815, plus goa_status concern (identical to VPS4A). Review thesis: VPS4B is the AAA ATPase that binds inside CHMP2A-CHMP3 tubes and disassembles ESCRT-III on ATP hydrolysis (direct biochemistry, PMID:18687924), NOT an ESCRT-III subunit; notes state "PN projection to GO:0000815 should not be added for VPS4B." Review instead asserts GO:1904903 ESCRT III complex disassembly (verified) + GO:0016887. GOA has GO:0016236/0006914/0061738/0097352 but no GO:0000815 or GO:0000045, so projecting GO:0000815 as "more_specific_than_existing_goa" mislabels a contested new CC as a refinement.
PN story / NEW pressure: Mixed. "ESCRT-III activity modulator" placement matches the review; the microautophagy "ESCRT-III complex component" leaf miscategorizes VPS4B as a member. Accurate story captured by GO:1904903/GO:0016887; no new term needed. (Review keeps the GOA late-endosomal-microautophagy row as UNDECIDED, autophagy/autophagosome-maturation/macroautophagy as MARK_AS_OVER_ANNOTATED.)
Mapping strategy: Change recommended (same as VPS4A). Microautophagy-component leaf→GO:0000815 should be context_only/no_mapping for VPS4B (consistent with the sibling modulator leaf), as VPS4B is not an ESCRT-III subunit. GO:0000815 over-reaches per the rejected broader-projection precedent; GO:0000045 also unsupported (review does not add it).
Evidence alignment: Partial. PN cites the MDPI closure review + "CHMP2A as a regulator of phagophore closure" (Nat Commun). Review draws on VPS4B-specific primary work (PMID:18687924, 18606141, 15024011, 16757520, 20616062, 24105262) not in PN.
Verdict: ESCRT-III-modulator framing consistent; GO:0000815 projection contradicts the review (VPS4B disassembles, not member). Recommended edits: none to YAML; reclassify PN microautophagy leaf→GO:0000815 to context_only for VPS4B and drop/relabel the GO:0000815 and GO:0000045 projections.