Hours later the statistics are correct.
I see that this all is eventually consistent in highly distributed systems, but it’s wild to have non-matching numbers in one place.
They could just have a JavaScript on the page sum up the bucket sizes to avoid mismatches.