Copied from the support board:

I have spectral libraries that return an error: "n runs were not converted due to insufficient correlation" when I try to import them into a iRT database. On investigation I found this occurs where one or two iRT peptides have no spectra in the library. It is possible to add the peptide in the spectral library to a document's iRT database via a roundabout route of importing them into a calculator based on the subset of iRT peptides present in the spectral library, finding the peptides in the swath maps, then importing them as results back into an iRT calculator with the full number of iRT peptides. Whilst it is not ideal to calculate iRT scores from a library missing some iRT peptides, it doesn't seem to make much difference to the regression, particularly when the missing iRT peptide is in the middle of the RT range. It would be nice if the "Edit iRT Calculator" dialogue could calculate a regression based on the iRT peptides found in the spectral library and import them, perhaps with a warning and an accept/reject option, rather than reject the library outright.

And a second post from a different person:

I also completely agree on the iRT calc. issue. I normally transform a DDA recording from a DIA sample into a strectral lib and from this lib I also calc. an iRT database and both are used to stir the DIA import. But sometimes, iRT peptides do not get picked in the DDA run (its shotgun!), or the resulting iRT spectra are below the sig. cutoff of the spectral matching. In any case, skyline will complain because, one or several of the iRT peptides are missing in the final lib. Since, the full iRT set is 11 peptides these missing values usually do not harm the regression. My current workaround is to manually check the spectral lib, which peptides are missing and to delete them from the list of std. peptides used by the calculator. This is of course suboptimal, because skyline could simply warn and still do the regression. A diagnostic plot of the regression incl. the possibility accept/reject peptides and missing values for the final iRT calc. would be a super cool feature for DIA users and save a lot of time.

We should definitely improve this so that people do not have to jump through such hoops for valid data within the normal range of expectations for iRT standards in DDA runs.

I think your fixes to add more tolerance in how we deal with iRT regressions should fix this issue.