I'm still looking for data, but here is my latest update in terms of totals, as well as some discrepancies / oddities I've identifed to date.
Total ballots cast (still counting): 132,159,642
Obama: 69,452,491
McCain: 59,929,852
Third Party + Write-in: 1,914,857
Other (total votes cast minus the first three): 862,442
Obama margin: 9,522,639
Margin %: 7.2% (52.55% to 45.35%)
In the "Other" category, there are possibly write-ins and third party candidates where those are not specifically identified by the county or state. I haven't kept track of which states / counties split out write-ins, but I would roughly guesstimate about 100-200k of the 862k were votes and the balance are under/over votes.
The only real change I expect now in totals for Obama and McCain are from Queens, NY. I give it a slightly better than 50% chance they push Obama above 69.5 million.
Now for the fun part - what I've found that's wrong. I'm in no way trying to start a conspiracy theory here. I think most, if not all, of these errors are just honest clerical errors. But I do think that it is important to identify these errors and request they be fixed. I also think it re-iterates how inefficient and prone to manual error our current election system is and the fact that we need to improve it to remove some of these opportunities for human error.
Here is what I've found so far and any resolution:
Broadwater, MT: The official state results submitted by the county did not add in a portion of the actual votes. My totals include the correct total. The individual I contacted at the county was more than happy to provide the correct results and was very apologetic about the error.
Nebraska: 4 counties showed total votes cast that did not make sense compared to the totals for President. 3 counties (Dawson, Merrick, and Nance) showed a very high rate of undervotes. Holt county showed more votes cast than turnout. I contacted the SOS, and she indicated the total turnout listed was incorrect, and they have updated their canvass book to the correct totals. The numbers now make sense.
Illinois: Will county on the state site is higher than on the county site, even though the county site is listed as official. St. Clair county is lower than the totals for St. Clair and East St. Louis, although the East St. Louis website is not listed as official. The turnout number for Will also seems amazingly high and would indicate close to 40k undervotes. I sent an email to both the state and East St. Louis county tonight. Update: Just heard back from them. Will's numbers don't include the City of Aurora, so they need to be added. And the East St. Louis unofficial numbers dropped when the official numbers were published. I asked if they knew why -- waiting to hear back.
Vermont: It appears 5 towns in Vermont did not include their absentee ballots in the total turnout. I emailed them last night and have not heard back.
Maine: It appears 16 towns in Maine have more votes for President than votes cast. The most egregious is Bucksport, with 715 votes cast and 2,572 votes for President. I also emailed them last night and have not heard back.
Maryland: Prince George's county is reporting about 2x the number of votes cast on absentee ballots when compared to votes for President. Interestingly, the state website showing total ballots cast has a number that makes a lot more sense than the county website. I emailed the county last week, but no response as of yet. My totals listed above include the state turnout number, not the county number.
I think I've emailed a couple of other counties for results to get the official results, but these are the biggest errors I've discovered to date. I'll add any additional findings as I have them, as well as any resolution to the ones listed above.
If you have questions or requests, please post or email me. The data is there, so if you're interested in certain cuts of it, it should be pretty easy to do.
As always, thanks for reading!