1. If the set of peptides in your libraries overlap with each other, and you are using these libraries to help Skyline decide which transitions to use, then it would be best to combine all of the libraries into one .blib file.
When Skyline is trying to decide which transitions to automatically give you, Skyline looks through your libraries and uses the first library which has a spectrum for that peptide. If all of your libraries have spectra for that peptide, Skyline will not necessarily be choosing the best spectrum.
When you combine your libraries into one .blib file, the BlibFilter algorithm looks through all of the available spectra for each peptide and chooses the most representative one.
2. If you want to compare results between groups of replicates, you should take a look at the Group Comparison tutorial:
If you have multiple samples in each cohort, and each of those samples has multiple technical replicates, Skyline does end up averaging the numbers for the technical replicates, and the fold change is calculated by doing a T-test on those average values.