This is a known issue I just haven't managed to release a patch for in Skyline 19.1. It is fixed in Skyline-daily. Skyline 19.1 works the same as Skyline 4.2 if you keep it from finding the source spectrum files and then choose to use the "embedded spectra" inside the msms.txt files. Skyline 19.1 was the first time we attempted to circumvent the issue with MaxQuant msms.txt spectra when MaxQuant is allowed to do charge state deconvolution, which means all peaks identified as belonging to charge 2 and greater ions are mapped and summed at the charge 1 m/z for the matching fragment. This can be turned off in MaxQuant, but I believe you also need to turn off deisotoping, which means you also get full isotope distributions in your MS/MS spectra and this I am told makes spectrum matching perform worse.
Anyway, in 19.1 if you keep Skyline from using external MS/MS spectra then you will get 4.2 behavior. In Skyline-daily which we hope will soon become 19.2, you should be able to use external spectra as we had intended in 19.1.
Sorry I didn't get out a 19.1 patch to fix this when we first realized it was an issue and fixed it in Skyline-daily.
Thanks for reporting the issue. Apologies for the time lost to understanding what was going on.