Match against the 519 named indicators, not all 689

170 of the 689 base indicators carry no name in any source we hold: they return neither metadata nor observations. The inverse crosswalk originally fell back to the DCID mnemonic, so a quarter of "the framework" was represented by query strings like DI ILL OUT.

Nothing can match that. It inflated the tail — datasets looked further from the framework than they were — and put mnemonics in the nearest-indicator column, where VC VAW SXVLN was reported as a cluster's closest SDG concept. Excluding them moved every weak positive control up by 4–16 percentile points.

The count of what was excluded is printed on the page. An analysis that quietly matches against garbage text and reports the resulting distance as a finding is measuring its own inputs.