I just replicated this on a brand new laptop, also running Windows 7 64-bit, but with an entirely different set of hardware. Once again, it seemed partly random. I extracted cmpro, copied the MESS and MAME executables directly into its root folder, and created the profiles. I created MAME first, which showed a "correct" screen with 14029/14029. Without restarting cmpro, I went back to the profiler and created the MESS profile as before, and it showed a Set Information screen like in my last screenshot: 1568/1647. Since this can take a few tries, I repeated the process.
I selected both profiles, right clicked and chose the delete option. Without ever restarting cmpro, I did everything once again, though this time I created MESS first. It returned 1568/1647 once more, and this time when I then created MAME the 14015/14027 scenario occurred.
Other filters seem to work fine, as I was able to do something like %f=vasara to select the correct set. Unfortunately, exporting the dat yielded the same inconsistent result-- sometimes it would load fine, other times not. As an alternative, I am uploading an example of an entire affected cmpro directory. As far as I can tell, extracting from this archive and running the program without touching anything will always show this issue.
http://www.megaupload.com/?d=R37942R4