Could be a lot higher: Democrats have won a bunch of seats in 2006 and 2008 that they really just have no business holding based on the PVI.
I think that this is really the key.
There are several factors.
First, the only time, post war that the president's party gained seats in an off year election was when the president had high poll numbers and it was the first year after a redistricting (2002). The most minor losses occurred when the president had higher numbers. My guess would be that even if Obama were at 50% to 55%, you'd see a 10-15 seat loss.
Second, there are a number of naturally GOP seats, that might include another 5-15 seats.
Two other factor will be if Obama is below 50% and if there is a strong anti-incumbency feeling by the election.
Naturally, I'd say 25-35 seats could be expected, though not significant. I'd say anything above 35, and you are entering re-alignment range. The GOP doing better than 1994 is
not out of the question.