Thanks, Brett. Matt, Brian, and I worked hard on the performance.
For my recent processing, we have used MaxQuant msms.txt files for the libraries. When doing this, I think you want to try to disable its fragment ion charge state deconvolution, or you will end up with MS/MS spectra with charge 1 and 2 fragments merged into charge 1 making them less than ideal for subsequent DIA library use, though, not completely unusable. If you have other search tools you would like to try this with, ProteoWizard msConvert should make that possible with exported MGF or mzML. I have heard PEAKS also works. We would be interested in working with you to support any other pipeline you have, if the search provider is also willing to do the work to pass through the IM values.
We have also tested a Mascot pipeline that passes through IMS values, and, yes, they are important. Once you have your library built, open it with View > Spectral Libraries. Review some of the spectrum matches and make sure you have both RT: #.### and IM: #.###Vs/cm^2
Then you want to go into Peptide Settings > Prediction (sorry, we are going to move this to Transition Settings > Ion Mobility soon) and check "Use spectral library ion mobility values when present". In my most recent optimization analysis, I came away feeling that a "Resolving power" setting of 50 worked best for the data I have, but anyway from 20-50 should work okay.
Thanks for posting on the support board. Looking forward to hearing more as you get going with diaPASEF in Skyline.