"1) There is almost always undervote in lower profile races. People go in to vote for their high profile guy and leave the rest of the ballot blank (explaining why the senate primary total has more than the AG race)."
True, but it seemed like the spread was larger in Delco. Maybe instead of a spread of 10,000 it should have been 5,000 or so, which could mean that there was a mistabulation of 5,000 votes. Maybe there's nothing wroing with the Delco results, but I'd give them a closer look if I were Toomey, especially since the GOP machine in Delco (the "War Board" I believe it's called) is probably not much different from the one in Bucks County.
"2) Castor did well in Deleware because it's in the same media market/region of the state that he is a very sucessful district attorney. The Castor/Corbett campaign wasn't really a left-right battle as much as a 'successful guy who can win the general'-'has-been guy that got the party endorsement' battle."
Well, it appeared to be a right-left battle in some circles, but you're probably right that Delaware County's location helped Castor. But if Castor was so popular in Delco because he is a next-door neighbor, why did 10,000 Republicans vote in the Senate primary and not the AG primary? I'd still look into the Delco results.
"3) The ratios you site. Sure about the registration totals for Westmoreland?"
I stand corrected about Westmoreland's voter registration. While it voted for Bush in 2000, it still has close to 1.7 registered Dems for every registered Republican, a lower ratio of Republicans than most if not all of the Western PA counties I mentioned. See
http://www.dos.state.pa.us/bcel/cwp/view.asp?a=1099&q=443600 , which numbers are from April of 2004, so they should catch most of the Westmoreland conservatives who have switched from Dem to Republican as the Dems move further left. But adjusting for the county's GOP registration numbers, GOP turnout still seems a bit low in Westmoreland, although not as low as I originally thought, and I'd still look into it.
"Also, there were other down ballot races that, if on the D side ,could have boosted the ratio in favor of Democrats in that area. Also, one of the Dem AGs may be particularly popular there."
Those are certainly possibilities, although the Dem AG race was split 36.6%-36.4%-26.9% among the three candidates, and there weren't any contested congressional primaries in the county (other than the Dem primary in the portion of the county lying in the 4th CD, but very few people voted in that one). You may be right and I might be making mountains out of molehills, but I've seen too many tabulation errors in other elections (heck, in this one too, in Bucks County) to believe that those results shouldn't be scrutinized.