I said I wasn't going to spend time on this, but I figured it out in the shower.
The 8: Harem Counting common event counts up everybody who has their portrait display turned on in the CSCA Harem script code and sets variable 5: Harem Count to that number. It gets called every time you see the Harem interface from Yarra.
Clearly, Ginasta's leaving the harem was never handled properly as far as CSCA Harem was concerned. Sure enough, cruise on over to 21: The Final Battle/Battlefield Locus (4,0) and there's a commented out call to function of CSCA Harem that doesn't exist and a note that Sierra can't be dealing with it. Apparently whoever originally implemented the script, presumably for Harem Collector, never considered the possibility that a girl would want to leave the polycule. (It looks like the thing was adapted for some kind of Pokemon Pokepedia type thing.)
Implementing the function was pretty simple, but somebody else can test it.
https://www.mediafire.com/file/1kl6ucextcqb0jl/RevisedCSCAHarem.txt/file
Just for safety, an invocation of 8: Harem Counting should be added to the end of Renthnor Wilds/Westbrook Village (10,2) to make sure the Harem Count doesn't get nerfed to 37 when Ginasta rejoins. Having the higher value is probably meaningless since I don't think there are any meaningful checks of Harem Counting in the Postgame.
Except...
Oh FUCK my life...
There's no handling for Neranda, Altina or Varia being dead.
UGGGGGGGHHHHHHHHHHHHH....