unexplained cuffdiff outputs for stranded libraries

[capricy]

Dear Community,

I recently had sequencing data from stranded library (second strand). The count output from STAR indicated that most of the reads mapped to the second strand. I then ran cuffdiff using the bam files from STAR with -library-type parameter specified.

I happened to find that the either -library-type fr-firststrand or -library-type fr-secondstrand give the fpkms. How should I explain this?

For example:

0610009B22Rik, has 131 reads mapped to the second strand, and only 1 read mapped to the first strand (from STAR count file)

however, it has fpkm of 10.05 when I used "-library-type fr-firststrand", but fpkm of 2.32 when I used "-library-type fr-secondstrand". My expectation was that the latter should be much higher than the former...

Could anyone provide some inputs?

Thanks a lot!

C.