different results in samtools flagstat and bowtie

[heso]

Hi,

I'm mapping human single-end miRNA reads to the genome using bowtie -v0 -k5
The reads are in fasta format and collapsed with fastX_collapser

So, for a stats output from bowtie I get:
# reads processed: 203607
# reads with at least one reported alignment: 85470 (41.98%)
# reads that failed to align: 118137 (58.02%)


after performing samtools view -h -F4 for the bowtie output --sam file
I get in samtools flagstat the following:
262622 + 0 in total (QC-passed reads + QC-failed reads)
0 + 0 duplicates
144485 + 0 mapped (55.02%:-nan%)

So there is >10% difference in mapping.Why?

I read somewhere that bowtie counts the nr of mapped reads as opposed to flagstat that counts the alignments. As I've got collapsed fasta file, wouldn't these be the same?Which one should I use when referring to the % of mapped reads?

In addition:
The mapping I get with bowtie -v0 -k5 is very low. Should I change some parameters?
I tried -v1 -k5 and got a higher mapping rate (76,55% from bowtie stats) , but isn't one mismatch too much to allow for collapsed reads?

I would appreciate any feedback...

[dpryan]

Well you're allowing up to 5 reported alignments per read, so that's enough to account for up to a 5x difference in what's reported (samtools is telling you the number of alignments, bowtie the number of reads producing them).