ORG full-year equivalence is a proxy, not a concept identity: claims + validation need alignment (#539 follow-up)

Post-merge audit finding on #539 (6e462ca), cross-family (sol adversarial pass with survey-definition citations). Severity: HIGH, concept integrity (no-fabricated-mechanisms class).

Defect

The transform computes annual wage income × 52 / weeks_worked and describes the result as matching an "annualized reference week" donor concept (org_wages.py ~:591/:603; source_stages.json ~:1517). That equivalence does not follow from the survey definitions:

• ORG PTERNWA is usual weekly earnings at the main job (multiple-job holders: main job only — BLS CPS definitions).
• ASEC WSAL_VAL aggregates wage/salary earnings across employers; WKSWORK counts weeks with any work, including paid leave (ASEC 2024 data dictionary).
• So the claimed equivalence holds only for stable, single-wage-job workers; it is an approximation for multiple-job, mixed wage/self-employment, and changing-earnings cases.
• The PUF channel is weaker still: PUF tax-unit income totals are allocated over cloned ASEC people while ASEC weeks are retained (puf_support.py ~:654/:1066/:1678); dividing allocated PUF income by cloned ASEC weeks has no demonstrated work-duration interpretation.

What held up under audit (keep): masked division handles zero/missing weeks; donor zero-hours filtered; no calibration-target coupling (pre-calibration feature path only); QRF seeding explicit; suite green on head.

Required

1. Rewrite the exact/concept-clean claims as what the transform is: a proxy with a stated validity domain (docstrings, stage description, PR-body-derived docs).
2. Add validation/tests for the semantic cases the claim glosses: multiple-job holders, mixed wage/SE income, PUF-supported records; plus the untested clipping edges (negative, >52 weeks).
3. Or: source a genuinely aligned recipient-side measure and make the equivalence real.

Owner: the org-wage-weeks lane (#529 / #539 authors). Related operational finding filed separately (materializer version bump).

Read the transform and claims in org_wages.py around lines 591 and 603, then compare its stage description in source_stages.json around line 1517. Inspect the PUF allocation paths in puf_support.py around lines 654, 1066, and 1678. Done means the documentation calls the result a proxy with a validity domain and validation covers multiple-job, mixed wage/self-employment, PUF-supported, and clipping cases.