ME Senate — Collins retention

Thesis

1. 1.PROBABILITY REVISION TIMESTAMPED 2026-05-22: EHIQ 62% → 50% after SEM v1.2 build + TIQ 2026-05-22 re-scoring. **LARGEST REVISION IN THIS UPDATE — FLAG FOR ERIC REVIEW.** SEM v1.2 me-senate-general probability = 25.34% (v1.1 was 25.43%; v1.2 mechanisms did not fire). Polymarket "Maine Senate Election Winner" — Republican (Collins) ~27% (per SEM provenance). Apples-to-apples gap: SEM-EHIQ = -36.7pp; Market-EHIQ = -35pp. BOTH disagree materially with prior 62%. Underlying poll Emerson Mar 21-23 2026 (n=1,075, MOE ±2.9pp): Platner 48 / Collins 41. Mills withdrew clearing D path to Platner. TIQ specificity Collins 30.59 (HIGH_CONVICTION per 2026-05-22 scoring) directly contradicts the prior EHIQ thesis claim of "highest desperation signal" — TIQ reads HIGH conviction (not weakness). §4 OOS Phase C FAIL (ρ=+0.18, p=0.35) prohibits claiming TIQ is method-validated, but the prior "desperation" reading was itself TIQ-research-internal. Conservative damping: not converged to SEM 25%/market 27% — preserved at 50% to acknowledge (a) Collins-specific moderate-R brand strength in ME, (b) Emerson poll is 2 months stale, (c) Dem primary 2026-06-09 still pending. Range 42-58% to reflect the material uncertainty band.