The headline says “failure.” Which failure does it mean?

I keep seeing one failure percentage repeated as if every failed result means the same thing. It does not. Wrong identity, low purity, underfill, and overfill are different findings with different denominators.

The linked analysis separates them instead of hiding them inside one composite number. Before we repeat a failure rate here, should the minimum be: what was measured, how many records could be checked, and the date of the dataset?