Separating real economic settlement from speculative churn is the hardest, most contestable step, so the desk fixes the rules before it sees the answer. Four statements are kept strictly separate: observed on-chain activity, then structural classification by output structure and provenance, then a bounded inference of defensible settlement, then any market overlay as display only. Classification depends only on structure, provenance, ownership, and versioned labels — never on a historical median, a target center, or a desired claim state. What cannot be attributed is not hidden; it is disclosed as an unknown share. The honest output is a range — lower, working, upper — never a single point, and it stays uncalibrated until the rules pass their gates.