Usage of HMM, in my opinion, is justified if you have an idea what the states should be. How do you imagine the states? 1st state - background rate, 2nd state - increased rate? I believe 1) the signal there is sparse, so you don't really need HMM, sliding window will be enough, 2) different cells have different mutational rate relative to their chromatin organisation (if we talk about somatic context - and as far as I know this "frequency of passengers" is most often used in cancer genomics). If we speak about germline - calculate everything in a windows of 1MB, make a histogram, classify windows accordingly to your purposes.