Sorry for being slow here... For some reason, applying the
transformation from the README file does not seem to be
appropriate for the dataset. I'm not quite sure why that would
be.
If there is any problem with it, then Mike's suggestion of
using 3dWarp would be a good way to go. We have used
3dWarp for alignment to a template, which may also correct
field distortions to some degree (your milage may vary).