It seems like ohio_anon and DCal in comments have figured out the mystery in OH-15. You've probably noticed by now that sites which rely on data from the AP (such as CNN) are reporting results that look like this:
Steve Stivers (R): 48% (146,907)
Mary Jo Kilroy (D): 43% (134,492)
Yet the race hasn't been called, and news accounts say Stivers has just a 321-vote lead. So why is the AP data screwed up? Well, ohio_anon points out that in Ohio, the board of elections for the county with the greatest number of people in any given congressional district is responsible for reporting for all the counties covered by that CD. This means that the data on the Franklin County website includes the votes for much smaller Madison and Union Counties.
So the AP appears to be counting Madison and Union twice - once on their own, and once folded into the combined results reported by Franklin. (Stivers did much better in the two smaller counties, while Kilroy ran ahead in Franklin, so that's why his lead is inflated.) Indeed, the math works out perfectly, as DCal observes - double Madison and Union and add them to the "Franklin" results, and you get those phony numbers listed in the blockquote above.
The AP needs to get its act together, but the bottom line is that due to the closeness of this race, we probably won't know the final results for ten days.
