I would suggest rereading the paper. The iRT method depends on the prior measurement of a peptide form on chromatography like what you will use in your experiment. (emphasis on like, because using an iRT library from someone else's chromatography has been shown to be less accurate than one on the same system you will use in your experiment) The iRT method is designed to distinguish between unmodified and modified forms, but these are not even always resolvable in the RT dimension, making the use of site distinguishing fragment ions necessary to measure some isobaric modified forms (https://www.ncbi.nlm.nih.gov/pubmed/28604659).
However, the accuracy of the method depends mostly on the variance of the form in your retention time measurements and the variance in your measurements of your chosen iRT standard peptides. If you could measure your iRT standards with the exact same relative separation every time (shifts earlier and later don't matter - if they all shift the same amount) and your peptide of interest exactly the same relative distance from them every time, then your predictions using iRT would be perfect every time.
To know how well iRT will work for your PTM studies, you should be interested in:
- Do your targeted transitions distinguish the modified form of interest from other forms of the same peptide?
- Can your peptide form be otherwise resolved in retention time by other forms that would be measured by your chosen transitions?
- What is the expected error in your RT prediction? (think residuals in a linear regression between predicted and measured)
But, if you feel confident that you have #1 solved, then the only reason iRT would be any less accurate for a modified form than an unmodified form would be that the modified form somehow varies more in the RT dimension in your measurments.
Hope this helps. People definitely use iRT in PTM studies.