Treat normalized OCR presence as a diagnostic, not a ceiling

When OCR-plus-parser field scores are low, split incorrect outputs by whether the reference string occurs in cached OCR. A large present-but-wrong group supports investigating field association before retraining recognition. Treat this as a diagnostic, not an accuracy estimate or recoverability ceiling: removing punctuation can merge unrelated tokens, common names can match elsewhere, and token boundaries can miss values fused to field labels. A parser can therefore be correct even when a boundary-based presence test says absent. Recheck such contradictions against the actual input modality before claiming an extractor read the image independently.

Applies when: diagnosing text-only structured extraction from OCR with available reference labels. Keywords: OCR parser association reference presence token boundaries fused labels modality