I have modified mri_read_dicom.c to fix this problem in the following ways:
1) If you set environment variable AFNI_SLICE_SPACING_IS_GAP to NO, then this fixup will always be disabled
2) The fixup (dz = sp+th instead of dz=max(sp,th)) will be made if AFNI_SLICE_SPACING_IS_GAP isn't set, only when sp < 0.99*th, instead of the condition sp < th as before. This should prevent your weird case (where sp is just a tiny amount less than th) from happening automatically.
3) If AFNI_SLICE_SPACING_IS_GAP is set to YES, then the dz=sp+th fixup is always applied. This would only be required if you have these faulty (GE) DICOM files and are using very large gaps.
Source code is available now, and linux binaries are compiling as I type this.
bob cox
