I have not read all the posts thoroughly but the gtx 295 is effectively two gtx 250's in a single unit. It's top single gpu speed is therefore limited to the speed of a gtx 250 gpu, however, its has twice as many gpu's.
So for things that parallelize well the 295 demo's well but for things that do not multithread well, then the 27x and 28x will out perform the 295.
"If I have seen further it is by standing on the shoulders of giants." Sir Isaac Newton, 1675