I've looked at the original images in freeview and they follow the correct talairach coordinates. I've run mri_convert again and the files follow the correct talairach coordinates. It appears this may have been a 'one-off' issue. Will update  after I autorecon everything.

- Bryan