Thanks for the quick reply!
1) I can see what you mean. The reason I was looking at the average is that there was better definition of the pons as there is more signal on the average (and less distortion), however the shape of the best SNR shape may match the individual directions better for an improved overlap/fit. I will look into this further.
2) For tortoise 3.0 do you know if we need IDL license in order to use the command line tools? Since I am going to be running the scripts on a grid computer system, I do not think that would necessarily be feasible. In any case, I have been getting very good results for eddy current and warping using AFNI tools (ie 3dAllineate/3dQwarp etc).
Thanks,
Ajay