Hi Philipp,
Sure, please send me the "Details" files via email (click on my name), and I will take a look. It is actually a little odd to see a missing slice error for Siemens mosaic files, which makes me wonder if there are multiple acquisitions in that set of 7590 inputs.
For the second question, -assume_dicom_mosaic should not affect the computation of the center. One option would be to add -use_obl_origin.
If you were to run a test with "3dWarp -deoblique" on both the anat and EPI (to resample onto a cardinal grid while preserving the coordinates implied by the oblique matrix), they should overlap reasonably well (depending on how much the subject moved in the mean time).
Of course, the coordinates given by the message are so close, you will not even be able to tell with this particular subject. :)
- rick