What did every inventory list miss? A committed capture-recapture procedure, with synthetic estimates and none from real tool lists

An organization that assembles a list from several sources, such as a cryptographic inventory for post-quantum migration or a communications archive, can count what the sources found, not what every source missed. Here the population scope, canonical-form rules, signed lists, observation mechanisms, an independence graph and the estimation specification are committed before counting. Lists binding one observation mechanism are typed as one list, and fewer than three lists are refused. For three lists a log-linear model with pairwise terms and no three-way interaction is applied. An empty observed cell yields an undefined determination, and a conditional association ratio past a precommitted bound yields a refusal naming the pair. A reviewer who authored none of the lists can re-derive the record offline. Synthetic examples estimate 31 missed inventory members (interval 16 to 67) and 34 missed messages (14 to 79) against a true 21. A shared outage draws a refusal. Real tool lists yielded no estimate. Three source scanners found 19 of 36 hand-labelled uses in one library. Their protocols counted three lists with empty cells, and a determination corrected after the runs types them as one list. Two runs then took static, network and binary lists over six TLS servers. In the first, the lists together missed 1 of 45 declared members, but the static and binary lists shared no member, so the determination was undefined. In the second, the binary list, scoped to each server's served key, held 6 members, all true, and the lists together missed none of 43 declared members, nor any of a second truth of 42 computed after the run. A key-file reader gave the static list every key the binary list held, so two cells stayed empty and the determination was again undefined. One operator produced every list, so these runs test counting, not independence.