[iskandr]

The sequence report now separates target-event positions from vaccine mismatches and differences between reconstructed windows. CD109’s KKK/FMV mismatch is visible even in W1; clicking a W label updates the vaccine-row highlights for that comparison.

• Mark RNA-call target codons and the region downstream of frameshifts, with ambiguity preserved. The panel has no annotated fusion breakpoints or novel-ORF boundaries, so none are invented.
• Star every most-supported window per sample/method, keeping ties and deduplicating RNA inputs. Add a filter and detailed per-peptide outcomes for leading-window disagreements; Exacto itself does not rank its outputs.
• Show the two amino-acid failures (ABI3BP M/W and ANKRD17 R/K) inline and make them filterable.
• Separate RNA evidence into 2+ mutant reads, single-read evidence, observed zero and unavailable counts. Use the maximum of Sid counts and distinct Exacto raw-read allele calls per benchmark library, never pooled counts. Keep both component counts and exclude corrected/assembled inputs from this threshold. Preserve overall scores and add the support-stratified breakdown.
• Keep full recorded peptide containment unchanged. Document CD109’s possible, unconfirmed KKK modification rather than silently removing it.

Validation: 142 Python tests passed (4 integration tests skipped), plus a passing regression for raw-RNA promotion; 45 JavaScript tests passed; ruff and site build passed. Browser checks cover CD109, VPS72 pairwise selection, both residue failures, support/disagreement filters, and a 390px mobile viewport. Benchmark inputs and archived scores are unchanged; no fresh Exacto run is needed.