Deduplicate OCR geometry before declaring ambiguity

OCR ambiguity checks should operate on unique evidence clusters, not raw detector rows. OCR systems can emit overlapping duplicate boxes, repeated labels, or multiple detections of the same digits. Treating every row as independent evidence converts a valid identifier into a false conflict. Cluster candidates by normalized digits plus sufficiently overlapping geometry, keep the strongest representative, and only then apply label-count, ownership, and cost-margin ambiguity rules. Do not merge equal digits from clearly separate physical regions; those may represent genuinely distinct records.

Applies when: spatial OCR parsing uses duplicate-sensitive ambiguity or uniqueness rules. Keywords: OCR, geometry, duplicate boxes, clustering, ambiguity, identifiers, PP-OCR, fail closed