Joe,
No, this does not necessarily do away with the obliquity warping. In some cases the datasets will be very far apart in terms of the centers of their datasets or for very large angles of oblique acqquisitions. For these cases, obliquing or deobliquing to match datasets with 3dWarp or even just aligning centers with @align_centers will be necessary. The script assumes a rough match between the two datasets to start, so if the datasets are somewhat close, then the align_epi_anat.py script should work well. If you do this, you should probably remove the obliquity info using 3drefit. At some point, I will add the obliquity transformations into this script, so it will be pretty much transparent.
For now, the talairach transformation of EPI data is done using the -tlrc_apar option to define the talairached version of the anatomical parent. You should notice that both the EPI datasets aligned to the anatomical orig view data and the tlrc view (defined by the tlrc_apar dataset) are created. @auto_tlrc has its own list of options I did not want to duplicate in this program at this point like the choice of template for instance, so that is left to the user to be done prior to running this script.
Ziad tells me it should make no difference whether you do the Talairach transformation before or after 3dDeconvolve as long as you are looking at just the beta coefficients. Other statistical parameters should not be interpolated, and it is better to do the interpolation on the original EPI data and recompute them.
There's also a problem I just noticed regarding writing out the first tlrc dataset, so I'll be updating this program to fix that tomorrow.
Regards,
Daniel