0 fpkm for genes with reads

[mparida]

Hi All
I have come across an issue of 0 fpkm even when there are reads for a locus.
I downloaded mm10 genome and annotation from https://ccb.jhu.edu/software/tophat/igenomes.shtml as recommended.
I am using TopHat v2.1.1 and Cufflinks v2.2.1 for this analysis.
I have 2 technical replicates that have almost similar amount of reads and correlate well when it comes to protein coding genes (R=0.99, pearson) but some genes such as follows:
Genename REP1 REP2
Snord104 0 81982.5
Show 0 fpkm in one replicate but huge fpkm is another replicate. This is confusing because the amount of read counts for this locus is close enough in both replicates (REP1=154 reads and REP2=199 reads).
I have read forums that say how this is a bam file and annotation/gtf file sorting issue and has been fixed in the latest version and yet I am getting results like this.
Please help me understand what I am doing wrong here.
I will be extremely grateful for your concern.
Here is my command:
REF=/home/mparida/MM10_GENOME/Mus_musculus/UCSC/mm10/Sequence/WholeGenomeFasta/genome.fa
ANNOTATION=/home/mparida/MM10_GENOME/Mus_musculus/UCSC/mm10/Annotation/Archives/archive-2015-07-17-14-33-26/Genes/genes.gtf
BOWTIE2IDX=/home/mparida/MM10_GENOME/Mus_musculus/UCSC/mm10/Sequence/Bowtie2Index/genome
REPEAT=/home/mparida/mm10_repeatmasker.gtf

cufflinks -o $ODIR/CUFFLINKS/ -p 16 -b $REF -u --library-type fr-unstranded -N --compatible-hits-norm -G $ANNOTATION -M $REPEAT $ODIR/Sample_2_REP1.bam
cufflinks -o $ODIR/CUFFLINKS/ -p 16 -b $REF -u --library-type fr-unstranded -N --compatible-hits-norm -G $ANNOTATION -M $REPEAT $ODIR/Sample_2_REP2.bam
Rocky

[mparida]

0 FPKM of genes with reads cufflinks

Hi All
I tried mapping my reads to the transcriptome instead of the genome and then computed the FPKM using eXpress algorithm (from the collaborating group that wrote cufflinks) and now my technical replicates are correlating well for protein coding and non-coding regions.
I also did a comparison between fpkm generated by cufflinks and eXpress for protein coding genes only and they correlate well (0.97). This analysis suggests that cufflinks is not very effective when generating fpkm for noncoding regions in my samples.
Hope this helps others.
Rocky