The single most important output is the noise floor, and the rule that follows from it. A baseline is not one number; it is a mean and a spread — the average of the repeated readings and how much they vary. The spread is the survey's own error bar, measured directly by re-occupying a hub on unchanged structure (Dives 3 and 4). From it comes the reading rule that governs every future comparison:
A later reading is treated as real change only if it differs from the baseline by more than the noise floor. A difference smaller than the spread cannot be distinguished from measurement noise.
So a result is never reported as a bare number to be read literally. It is reported as an index with its spread, against a baseline with its noise floor, and any claim of change is checked against that floor first. This is what keeps the survey honest about what it has actually detected.
Compare like with like. Two readings are compared directly only when they were taken under similar conditions. A reading made under conditions unlike the baseline's is still reported — but with that difference stated as a caveat, never silently folded into a claim of change. (Appendix C explains why: the survey is observational, so conditions are matched, not modelled.)