This way, we don't need to create a new Allele for every read/Haplotype pair to be placed in the PerReadAlleleLikelihoodMap (very inefficient). Also, now we can easily get the Haplotype associated with the best allele for a given read.