You can inspect the images to see what is going on:
 1.) The original T1-scan is saved
   20130204_17443618NSFUSBRAINPREOPs1000a1001.nii.gz
 2.) Presumably, this is a 3D image acquired in sagittal orientation. The data is losslessly rotated to the nearest orthogonal orientation to support viewers that ignore the spatial transformation matrix.
 o20130204_17443618NSFUSBRAINPREOPs1000a1001.nii.gz
 3.) The software detects that there are excess slices of the neck that could disrupt normalization, so a cropped version is created
   co20130204_17443618NSFUSBRAINPREOPs1000a1001.nii.gz

(1) and (2) are identical for tools that support the NIfTI spatial transforms. (3) often works better for tools that assume your image has roughly the same amount of neck as the template images. I am not familiar with BrainSuite, but if the output of (3) passes visual inspection I would think it is your best choice.

While I hope my old software is robust and mature, I would suggest you upgrade to dcm2niix (which comes with MRIcroGL).