Consistency: Internally consistent. Deep research (Falcon) and review YAML both center ATL3 on GTP-dependent homotypic ER membrane fusion / tubular-ER maintenance; both treat the ER-phagy receptor role as unsettled. The notes explicitly document the PN reticulophagy projection and the reasoned decision NOT to add it, quoting Bryce 2023 framing ER-phagy as "possible" context. No contradictions.
PN story / NEW pressure: PN asserts ATL3 as an ERphagy receptor → GO:0061709 reticulophagy, which is NOT in existing GOA and is a real term. The review correctly declined to ADD it: the ATL3-specific evidence supports ER fusion/morphology, and the GABARAP-interaction/ER-phagy receptor claim (Chen 2019, not cited in YAML) is not anchored to a cached supporting quote. Verdict: PN story over-reaches as a confident GO assertion; appropriately held as suggested question/experiment rather than a NEW annotation.
Mapping strategy: Gene does not change the node mapping; the type-leaf mapped/ok_for_propagation stands as a class-level statement, but this individual gene is a defensible decline (like TOMM20/HSPA8 precedent where projection out-reaches gene evidence). Projected term is narrower than review scope only in that review omits it entirely.
Evidence alignment: Divergent. PN cites review-style titles (Chen 2019 ATL3 ER-phagy receptor; Liu/atlastin-ULK1 JCB) not present in the review's PMID set. Review is built on fusion/structure papers (PMID:37102997, 27619977, 28602821, 24459106). No PMID overlap on the ER-phagy claim → the projection is unsupported within the dossier's own citations.
Verdict: Consistent; PN reticulophagy claim correctly NOT added (over-reaches gene-level evidence). No edits required.