I'm not sure why you would get such low Z.
Something that went wrong in the estimation of the MNI affine transformation with SPM function spm_maff8?
Create a new subject, import the MRI only and immediately compute the MNI transformation. Then check where the default AC/PC/NAS/LPA/RPA fiducials are placed. This would be a good indication of the failure of this registration method.
Example for a point in the right fusiform of the MRI used in the introduction tutorials.