# Proper Use Guide conversion QC

**Result:** PASS
**Generated:** 2026-09-15T17:00:47+09:00

## What was checked

- Every source PDF page has exactly one HTML page image.
- Page dimensions match the PDF CropBox, excluding printer crop marks outside the visible page.
- Every generated page was compared with an independent PyMuPDF render.
- Source hashes, page-image hashes, page anchors, duplicate IDs, and searchable text pages were verified.
- Conversion-only page captions are absent, so only page numbers printed in the source PDF remain visible.
- The PDF/HTML comparison page includes working Japanese and English controls.
- The separate Hybrid HTML Evaluation page keeps each approved page image intact and overlays selectable HTML only for headings and tables.
- Figures, charts, illustrations, QR codes, and other graphics remain embedded in the source-faithful page image and are not converted.
- All 48 hybrid page records were checked for valid normalized positions and source-matching heading/table text.
- The previously released Proper Use Guide library, readers, page images, PDFs, search indexes, and normalized manifest remain byte-identical; only the isolated evaluation route is added.
- The existing Package Insert engine, QC, generated documents, site source, deployed artifact, and source PDFs remain byte-identical to the protected baseline.

## Per-guide result

| Guide | Pages | Searchable characters | Lowest independent render similarity | Result |
|---|---:|---:|---:|---|
| トルカプ 適正使用ガイド (`trq01`) | 28 | 84,967 | 0.9934 | PASS |
| エトカマ 適正使用ガイド (`etc01`) | 20 | 49,883 | 0.9916 | PASS |

## Interpretation

The visible HTML uses high-resolution, lossless page images as the visual authority. This preserves complex tables, charts, checklists, annotations, QR codes, side tabs, typography, and branding without reconstructing or reflowing them.

The isolated hybrid layer exposes headings and detected table content as selectable HTML while deliberately leaving body copy and graphics in the source-faithful image. Table text is represented semantically for evaluation, while the visible table geometry remains the approved image.

The Hybrid HTML Evaluation page is a direct-link review aid. The source-faithful page image remains authoritative; reviewers can select and copy only headings and table content without changing the released reader.

The earlier source-checked contents/glossary pilot data is retained for traceability, but the current hybrid preview uses PDF-positioned heading and table regions consistently across all 48 pages.