The influence of data evaluation parameters on qualitative and quantitative results of untargeted shotgun profiling of enzymatic and nonenzymatic post-translational modifications (PTMs) was investigated in a model of bovine whey protein α-lactalbumin heated with lactose. Based on the same raw data, individual adjustments to the protein database and enzyme settings of PEAKS studio software increased the identification rate from 27 unmodified peptides to 48 and from 322 peptides in total to 535. The qualitative and quantitative reproducibility was also assessed based on 18 measurements of one sample across three batches. A total of 570 peptides were detected. While 89 peptides were identified in all measurements, the majority of peptides (161) were detected only once and mostly based on nonindicative spectra. The reproducibility of label-free quantification (LFQ) in six measurements of the same sample was similar after processing the data by either the PTM algorithm or the LFQ algorithm. In both cases, about one-third of the peptides showed a coefficient of variation of above 20%. However, the LFQ algorithm increased the number of quantified peptides from 75 to 179. Data are available at the PRIDE Archive with the data set identifier PXD050363.