Just wrapping up our off-list conversation for posterity here. Thanks for making your Skyline file available to me.
The issue is that pre-4.1 Skyline would allow any grouping of small molecule precursor ions under a parent molecule, each precursor having its own individual chemical formula that included ionization. As of 4.1, Skyline strictly models precursors as adducts applied to a common neutral molecule, which brings it into line with our proteomic model of neutral peptide with varying modifications and charges. Having this unified molecular model allows more of Skyline's capabilities for quantification etc to be applied to non-peptide molecules.
Unfortunately, Skyline is doing a poor job of finding the commonalities of your precursors and translating that to a neutral molecule with multiple adducts. It's a little better with the current state of the Skyline 4.1 code (patch release coming soon) - I can at least see the chromatograms - but I'm still not happy with the way Skyline is translating your document.
Your document is completely rational, and it's clear that, for example, your first entry, currently described as ions C20H31O5 and C20H27H'4O5, should be understood as neutral molecule C20H32O5 with adducts [M-H] and [M4H2-H], but Skyline isn't getting that right.
I would advise sticking with 3.7 for the moment, while I work to understand what Skyline is doing wrong. Josh is quite correct in saying that we value backward compatibility, and we'll make this right. Sorry for the bother!
Skyline Developer Team