I've been working with mouse data (but not resting state) for quite some time and unfortunately there is no standard brain, and we were not very successful with coregistering to T1 images. There may simply be not enough structure in the mouse brain (no sulci, gyri like in the human brain) to make meaningful transformations.
Although, I think some resting state people have been claiming to compare regions according to the Allen Institute mouse brain atlas using FSL for coregistration...

Why don't you just pick one EPI image as template and coregister all others to this image for group analysis? This works really well with linear registration using 3dAllineate, but the slices should be placed as similar as possible in each animal.
Then you could still manually draw ROIs according to an atlas, which is tedious but I would trust it more than some untested automatic coregistration...