Digesting (trypsin or whatever other proteolytic enzyme) proteins generates multiple peptides so the degree of complexity of the sample, at the peptide level, increases a lot. In addition there is the problem of infering the original protein from its constituent peptides. Why is this digestion step needed when you have to go back to protein level ? Is it just technological limitation?
This is going to be a very long answer but to give a short response.
You have to consider that MS for peptide detection works on the bases/principle of mass to charge (m/z) to detect an AA molecule, which is then normalised and analysed etc etc (http://en.wikipedia.org/wiki/Mass_spectrometry). Once you have the amino acids, then you just look at the order in which they get through and then you have your peptide sequence. Please read about the B, A and Y-type ions although this page is slightly technical (http://www.matrixscience.com/help/fragmentation_help.html). If the proteins was not digested then it would be far too big to be analysed since it would be read as one massive blob of m/z, which could be anything! So once you have your spectra from a peptide, you need to compare it to a model spectra, and based on that you predict what the sequence is as fragmentations peaks can be anything. Look at SEQUEST (http://en.wikipedia.org/wiki/List_of_mass_spectrometry_software). Now there are tons of correction algorithms that are applied, which you can look up.
Now RP-LC is carried out before MS to prevent all the sample rushing into the MS machine all at once as the machine can not cope if too much sample is inserted hence the flow rate is controlled. What I think you are thinking in terms of the set machine limit, is the (digested) peptide fragment that enters the machine, before undergoing electrospray. Now the experts that run the MS machines do set a peptide mass upper and lower limit, for what size/mass of fragment to accept and what size/mass to ignore. Look at this theoretical digest example (http://prospector.ucsf.edu/prospector/cgi-bin/msform.cgi?form=msdigest). This happens frequently with trypsin digest as it recognises specific AA before cleavage (http://en.wikipedia.org/wiki/Trypsin). Now a days, MS has advanced so much that we have machines such as Orbitrap (http://en.wikipedia.org/wiki/Orbitrap), which can cope with peptides, digested with elastase, which cleaves non specifically, which means you are less likely to get peptides that are far too big for analysis, hence increasing the protein coverage.
Hope this helps!
I found the answer to my own question It was formulated in a Nature publication by Mathias Mann, a pope in Proteomics: