If I'm understanding Nicholas correctly, the 10th percentile is based on individual samples from all testing sites across the country, which would still provide a functional "zero" baseline. For example a local site routinely reports zero virus, whereas compiled data will probably never show zero.