After a Bell-test quantum randomness run, what is the point of compressing millions of imperfect raw detector bits into a much shorter public string?

Answer: To reduce bias into cleaner bits

To conceal detector flaws — Concealing detector flaws would undermine the point of a certified randomness experiment. The apparatus and assumptions need to be analyzed openly enough to justify how much usable entropy was produced. Compression is not a cover-up; it is a way to keep only the amount of randomness the evidence can support.

To reduce bias into cleaner bits ✓ — Raw quantum measurement strings can be random yet still biased or imperfect. A randomness extractor compresses a long raw string into a shorter string that is much closer to uniform. That is the opposite of the everyday intuition that keeping every bit must be best: for certification, fewer bits can mean cleaner bits.

To prove raw bits were uniform — Compression does not prove the original raw bits were already uniform. If they were perfectly clean, there would be less reason to throw so many away. The extractor works because the experiment can certify enough entropy in the raw data, then distill that entropy into a shorter public result.