Thanks Nemo,
Those stats work on live edits in article space, hence the figures of about 3.5 million a month in the UK as opposed to the 5 million revision IDs.
The ten million interval data is calculated from the rawest of raw data, the actual revision IDs. They started at 1 in January 2002, so it is possible that numberofedits is partly different due to the first years edits, but as the next million took 17 months I'm not expecting that the first year was a million, let alone 100 million.
If {{NUMBEROFEDITS}} is just corrupt data then that's unfortunate, but if it were for example being incremented for each edit saved or otherwise then that would give us a count for edit conflicts, so it would be good to find out what it actually measures.