the successful ones are easy, but the others are more difficult/murky/interesting, and typically map to pg essays in comical fashion.
all the standard sources: their sites themselves, their blogs/twitter accounts, crunchbase (about same vandalism rate as wikipedia), linkedin, facebook, archive.org, hacker news, techcrunch, venturebeat interviews, school newspaper interviews, googlefu. i think the more important point is that where this is wrong, to the extent that people care, they are going to let us all know and it will be fixed.
Although a quick look through your spreadsheet indicates you seem to have done a pretty comprehensive job !
I noticed you've got ModFour as a dev shop, they actually went through YC with a product called ShopKick (see my answer at http://www.quora.com/What-did-the-Y-Combinator-startup-Mod-F...).
I've also got a twitter feed of all the YC companies: http://twitter.com/#!/imranghory/ycombinator
(there are also ycombinator-wXX and ycombinator-sXX lists for the individual batches)
I agree; it would be very useful to understand where that came from. Though it looks like someone may have taken my original data above and gone through to look up articles/information to flesh everything out? (Because they kept my placeholders, where I knew how many YC companies were in a class though some hadn't launched.)