As you said, the SW is a bit bitchy in CW mode. It looks like, if I click on pnts first, CWfrequ 2nd, sweep time is around 250 ms. If I click the other way round (last click on pnts), the reference subtraction in the polar plot is not working anymore and the sweep time is indicated at around 450.
However, I figured that out early on that you need to do a click on CW frequ as last action in order to have CW working. I use your ref-sub feature for all my radar tests, and it messes completely up if you do not follow that rule.
When going to your raw data tab, it looks that clicking on "points" activates a frequ sweep mode, abandoning the CW mode previously selected.
(That is one of the reasons why I suggested earlier to indicate somehow if we are in "center span" "min/max" or "CW" mode)
Interestingly, the S21 data recorded seem to come with 5 ms interval. This may mean, that the 2.5 ms is for S11 and S21 combined