Okay,
The alliance member count figures are correct.
The history page has a bug (previously undiscovered) where, when you launch at a target, recall, relaunch at the same target, recall AGAIN, with the same fleet slot, all before the first fleet should have landed, it duplicates some of the entries. Apparently you've done this 4 times, hence the difference of 4. It was a bit hard to spot; I had to review your fleets with the two different queries / approaches to identify.
The planet stats on the beta history site are wrong in one way which I didn't really consider when I originally wrote the query:
I joined all the data by co-ords and so this kind of messes things up historically. I need to fix this!
To answer the question about different mechanisms - there's basically 1 driver table which logs this information but different ways of displaying the data. The aforementioned bug turns up because we are trying to get not only the launch data but also the recall data, whereas on the alliance member screen we don't care about recalls. For the beta history site, I'm calculating everything on the fly over all sorts of tables to get the data at once.
We don't store all the counts separately because 3rd normal form, and we're not running some sort of OLAP database
Edit:
p.s. you all suck, no one nudged me or even bumped this thread