Hi, Matt-
Glad the article was of interest!
I think the main thing happening is noting the distinct usage of 2 AFNI programs for adjusting header information and datasets:
+ 3dresample is useful when the current header information is *correct* (in the sense that the data on disk is consistent with the header info), and when you change a header property like orientation, the data also gets changed to stay correct.
+ 3drefit is useful when the current header information is *incorrect*, and you want to change a header property like orientation, and the data will *not* be changed simultaneously.
As a corollary, if your data *starts* correct and you change the header property of orientation with 3drefit, then it will *change* to be incorrect.
If I understand correctly, that is what has happened in your case?
--pt