Please send me the inputs off-list and I will take a look.  I suspect the problem is that the "original MTZ" isn't really the appropriate file - there is a weird corner case where the output of Aimless (I think) is recognized as "unmerged" even though nearly all reflections are unique.  This could explain the weird statistics, in which case my recommendation would be to re-run the scaling exactly as before but also directing Aimless to output the unmerged reflections as a Scalepack file.  (Or, in the future, if you process your data with xia2 it will do this automatically, but I don't think you need to re-process these data to get what you need.)