Consistency: Deep research ↔ review ↔ PN annotation consistent. LMAN2/VIP36 is correctly an L-type (ConA-like) lectin cargo/sorting receptor (D-mannose binding GO:0005537, carbohydrate binding GO:0030246) — NOT a mannosidase, matching the PN "Lectin chaperone" type. Distinctive feature well captured: retrograde Golgi→ER recycling (GO:0006890) in post-ER QC of alpha1-antitrypsin, plus a secondary shed-ectodomain/phagocytosis role kept non-core.
PN story / NEW pressure: No NEW GO pressure; functions are captured (IMP D-mannose binding, retrograde transport, ER-Golgi cycling). Falcon shedding cleavage-site detail (PMID:33796845) refines the already-non-core shedding role only. Conclude: already captured.
Mapping strategy: Same over-reach as LMAN1 — group→GO:0006487 over-reaches for VIP36, a glycan-reader/sorter that neither installs nor trims N-glycans. new_to_goa would introduce an unsupported biosynthesis assertion. The review's own BP terms (GO:0006888, GO:0006890) are transport/QC, conceptually disjoint from GO:0006487. PN node correctly leaves "Lectin chaperone" unmapped; the inherited group term should not propagate.
Evidence alignment: PN dossier is mapping-only (no titles). Review evidence PubMed-verified and centered on PMID:20477988 (the key functional paper), 23701871, 22016386, 15308636. No bibliographic conflict; divergence is conceptual (group term vs. lectin function).
Verdict: Review strong and PN-consistent (lectin, not mannosidase; distinct retrograde-QC angle). Inherited group→GO:0006487 projection is an over-reach for VIP36. Recommended edits: [MAP] Do not propagate GO:0006487 (protein N-linked glycosylation) to LMAN2/VIP36 — it is a high-mannose glycan-reading sorting receptor (transport/retrograde QC), not an N-linked-glycosylation enzyme; flag the "N-glycosylation system" group node as over-broad.