Correct interpretation of tandem mass spectrom- etry (MS/MS) data is a critical step in the protein identifi- cation process. Comparing experimental spectra against a library of simulated spectra generated from a database is the most common strategy for this interpretation. Unfortunately, problems arise when treating unsequenced species since, in this case, the proteins to be identified are absent from the databanks and experimental spectra can only be compared to theoretical spectra from close and already sequenced organisms. In this context, spectra comparisons become a notoriously difficult problem. In this paper, we deal with this problem by considerably improving PacketSpectralAlignment ( PSA ), a method we presented in [1]. First, we explain how to take full advantage of PSA by carefully selecting the most promising alignment positions during the algorithm, and how to precisely fix the parameters of PSA . Second, we present a new method, referred to as PSAwEL , which allows a better localisation of modifications. We then propose a new peptide identification framework that integrates these improvements. Finally, we propose a comparison between PSA and the reference, SpectralAlignment [2], which shows that PSA behaves better in terms of: (i) quality of the results; and (ii) execution time. Our tests were conducted on the ISB dataset [3]. We then validate our new framework on Brachypodi