caveats

The four things that must accompany any display of this verdict.

These are shown to an analyst looking at their own sample, so they say what the verdict is and what it is not, and report no outcome of the experiment behind it. The rates the third caveat used to quote — recovery reaching 98% where the catalog holds the truth's families and falling where it does not — are findings about the method over synthetic truths, not evidence about the reader's data, and they belong in the project report. What survives is the part an analyst can act on: that the direction depends on something they cannot observe.