If the 18% of "republicans" that voted for Coons had instead ALL voted for O'Donnell, she still would have lost.
Here is the actual table from the CNN exit polls (sometimes people talk about a different exit poll when they suggest O'Donnell won the independent vote): CNN Exit Poll Delaware:
Final counts: Total: 296924; Coons: 173,900; O'Donnell: 123,025.Total Coons O'Donnell Other/No Answer Democrat (44%) 89% 9% 2% Republican (30%) 16% 81% 3% Independent(27%) 48% 45% 7%
Republican vote 30%*296924=89,077.2; 16% of that vote=14,252.
So if we added the 14252 to O'Donnell's total and took them from Coons, the final would be Coons: 159,624, O'Donnell 137,277. So she would still lose by a considerable amount (54% to 46%, an 8-point loss).
So you can't blame the republicans who showed up and voted for Coons instead of O'Donnell.
Further, it is silly to look at the actual voters on election night, and assert that the same voters would have shown up had Castle been the nominee. Especially when another argument made is that Castle voters stayed home, hurting O'Donnell. If the voters who showed up for a coons/O'Donnell race would have only barely chosen Coons over Castle, it is clear that in a real Coons/Castle race Castle would have trounced Coons.
Of course, that argument was clearly made when this exit poll was actually NEWS 2 months ago, and the poster of THAT thread ignored it.
Why this same exit poll is considered news by ANYBODY at this point is the more interesting question. Some people like to keep talking about the past hoping they can change it by spinning it.