What reference model do you align the RNAseq against? That would seem to have a huge effect on results no?