Expanded conference schedules shrink the number of pre-season games and, therefore, increase the relative importance of the pre-season slate even though it is fewer games (fewer data points) and earlier in the season. This serves to skew the power rankings due to greater error based on fewer data points heading into conference play. The distortion persists through the rest of the season until the post season tournament.
If the priority is to achieve more accurate power rankings and seedings in the tournament, then it would be better to allow teams about 5 pre-season scimmages that don't count. Then go directly into conference play. At the conclusion of conference play, construct out of conference schedues based on a round-robin of dual meets among the conferences (like the ACC-BIG, BIG-Pac-10, SEC-BigEast, etc. ) where the shedule is based on CURRENT conference standings. That way, every team's out of conference schedule will be against similarly ranked teams in ten other conferences. For example, If GT finished 5th in the ACC, then their out of conference slate would be playing the 5th place team from 10 other conferences.
Do that, and the power rankings would be pretty accurate heading into the tournament.