Hi Mike,
are you using the implicit mask setting? I'm not exactly sure what the problem is, but you can probably avoid it by generating an explicit group mask with the DPABI Quality Control menu. Using this option is preferable anyway since you want to have a group mask where at least 90 % of the voxels from all subject have some EPI signal in it. You can also take a look at the masking toolbox for spm:
http://www0.cs.ucl.ac.uk/staff/g.ridgway/masking/
Might not be an answer to your question, but hope this helps anyway ;)
greetings
David