We haven't disclosed the actual representation. You're right. It is a subset—opted-in users—of a subset—Android phone users and Google account users. What is useful from the information is the way that it provides a comparison with a baseline before COVID-19 hit, which is from early January to early February, which is relevant for most of the world. It provides that comparator, which is a data point that's useful as people try to assess whether or not they're seeing behaviour change.
On May 21st, 2020. See this statement in context.