Skyline does have an extraction mode with a mass tolerance when you choose the "mass analyzer" value "Centroided", which we highly recommend for both Thermo Orbitrap and Bruker TOF data. This will allow you to set a mass tolerance in PPM, and it has been my experience that this works best when you use the mass error mean bias + 3 SD as your tolerance, at least in the large-scale DIA runs I have analyzed.
This means, for instance, if your mean mass error were 0.5 ppm with a standard deviation of 1.5 ppm, you could expect:
0.5 + 3 * 1.5 = 5 ppm
A mass tolerance of 5 ppm would work well. If however, you saw 1 ppm mean mass error with a standard deviation of 3 ppm, then:
1 + 3 * 3 = 10 ppm
A mass tolerance of 10 ppm would be more advisable.
In most of the large-scale DIA proteomics data I have seen somewhere between 10 and 20 ppm works best, with 15+ ppm being a safer bet with new unoptimized data, since lower PPM values require more careful consideration, but certainly yield better results if your acquired data can support them. If however, say you have a +/- 5 ppm mean mass error and a standard deviation of 4 ppm:
5 + 3 * 4 = 17 ppm
Using a 10 ppm mass error with such data will definitely yield worse results than 17 to 25 ppm.
I know this is not exactly your case, but I would expect these findings to hold on a smaller targeted scale, and certainly it sounds like you want to switch to using the "Centroided" mass analyzer.
Let us know how it works for you. Thanks for posting your question to the Skyline support board.