The data (and you have to scroll down to find the details) is based on antibody tests carried out on a sample of 1000 people living in London and elsewhere and extrapolated from that, and not on people who have tested positive for the virus. The selection criteria are not mentioned, but a random sample of 1000 would have an error margin of 3%. False positive rate for the test is 0.2%: https://www.bbc.co.uk/news/health-52656808