According to a diary on the top of the Rec list right now, you should deduct 15% from Clinton's vote in IN because it is from Republicans who would not vote for her in the GE. However, the diary misread some of the exit polls, and acknowledged it. I posted the correct numbers in the diary, but nobody seems to have noticed it, so I'll redo them here.
One of the questions on the first page of the North Carolina and Indiana exit polls is who the respondent would vote for in a hypothetical McCain-Clinton matchup in November.
The result in Indiana was that 17% would vote for McCain. However, of those 17%, 41% voted for Clinton in the primary anyways. I think its safe to say that their support for the honorable Senator from New York is questionable at best if they wish to support McCain in the GE. 0.17*0.41= 6.97%
We need to balance out the effect from McCain-supporters who voted for Obama, as Seneca Doane rightfully notes. So, in an Obama-McCain race, 19% would vote for McCain, of which 12% voted for Obama in the primary. 0.19*0.12=2.3%, so the net gain for Clinton is 4.7%
Using the same methodology for NC, 15% of primary voters would vote for McCain in a McCain-Clinton GE fight. Interestingly, Clinton WINS that group of voters 46-43%, going to show how effective Limbaugh's efforts have been! 0.15*0.46=6.9%, about the same number as for Indiana.
19% would vote for McCain in an Obama-McCain race, of which 9% voted for Obama in the primary. 0.19*0.09=1.7%, so the net gain for Obama is 5.2%
So, Clinton gains around 5% from Republicans who aren't willing to support Democrats.
Sorry for the short diary, I usually like to have a lot more analysis than this
Update: Since I've already published this diary, I'll add some more analysis by looking at previous primaries. In PA, Clinton gained 3.1% from McCain supporters, while Obama gained 1.5%. So, she only had a net gain of 1.6%, not too significant of a factor. Unfortunately, it doesn't look like this question was asked for earlier primaries like OH and TX, so I can't make any analysis about those two.