It might have to do with the testing technology. If your false positive rate is high, but you’re only testing people with clear symptoms, it’s fine. But if you test random people the data could be too noisy