I agree but I don't think it's random error causing almost every test (RNA, antibody, etc...) that checks for prevalence to be skewed to much higher numbers of infected in the general public than most models predict.I don't know that his reasoning is sound as it depends on a number of variables that he seems to be completely ignoring.
There's the question of who is being tested, how are they being tested, what is the quality of the tests being used, just to mention a few.