Score dominant RNA windows and preserve native terminal lysines

[iskandr]

The vaccine report previously scored recovery across every output and marked native terminal lysines as solubility additions. It now scores the window with the largest raw or corrected RNA count in one sample/method, keeps all ties explicit, and retains weaker windows for diagnosis. A one-input alternative no longer penalizes a matching 20-input leader.

• Replace the dense header with a compact support table that separates mismatches, partial/unaligned outputs, mixed ties, and no output. Mark primary windows inline and make the disagreement filter follow them.
• Preserve native, parent-supported and encoded mRNA lysines. Exclude suspected terminal tags only with independent same-gene GENCODE context; freeze that evidence and bind it to scoring. Use KK in the legend.
• Re-score the saved Exacto 0.5.0a1 outputs: 17/36 primary recoveries; all-output diagnostic totals remain 25/36 targets and 57/78 entries. Every 10+ input leader recovers a peptide (12/12). ANKRD17 and tied TECPR1 have comparable primary disagreements, both at one input.
• Share primary ranking between the browser, analysis CLI and site build, refreshing it on every build. Document ABI3BP’s ATG/TGG reading-frame discrepancy.

Validation: 182 Python tests passed with four Linux integration tests skipped locally. All 52 JavaScript tests, lint and site build pass. Browser verification confirms native K preservation and dominant-window labeling. Frozen context regeneration is identical; archived protein sequences, RNA candidates and artifact hashes are unchanged. This reuses archived benchmark outputs; it does not rerun the full remote-data matrix.