The reported unreliability of anti body testing for those that had mild/no symptoms.
Things are NOT binary. Amounts matter. So if you have low levels of antibodies they might be below the limit of detection.
The test is specific for antibodies against one particular part of the virus. If you have made some against a different part, they won’t show either. We use the tests every day to see whether patients have made antibodies against the biological drugs we produce. The assays are not as sensitive as the similar ones we use to measure the amount of drug the subject has in their body.
Test for amount of virus, amount of drug, are much more specific because you design them to probe what you know you want to find, so can test for this. The other tests must look broader to what has been produced. You must make a choice based on what the majority seem to have made.
Numbers matter. For those US hydroxychloroquine doses. 29 million doses of 200mg or 29 million course of 1200+600mg/day for 10days (88000 treatments?). Which number would you release?
[TL:DR] tests always have a limit of detection and only look for what you think is important. That makes them fallible. And dose matters.