Why nothing is filtered out Shipped
Dropping low-confidence findings below a threshold is the industry default, and this pipeline refuses it — asserted by a test, not by a policy document.
Every finding the database holds appears in the report, ordered by what matters. The asymmetry is the argument: a missed expensive-to-unwind issue is the costly failure, and a low-priority note you skim past is not. A confidence threshold trades the cheap failure for the expensive one and calls it noise reduction.
Ordering does the work a filter would
The consultant's score is what puts the real defects at the top. A reader who stops halfway down has still read everything that mattered — which is the actual goal a threshold is reaching for, achieved without deciding on the reader's behalf what they were not allowed to see.