miraligner output reads

Hi,

I'm giving this tool a try for the smallRNA project we're working on as it adjusts perfectly to how we want to align reads to hairpin sequences to get mature miRNA counts. I have a question regarding number of annotated reads vs number of non mapped reads. I'm executing miraligner with collapsed reads (minimum length of reads is 16) and this parameters:

java -jar $miraligner -minl 16 -freq -pre -sub 1 -trim 3 -add 3 -s hsa -i $inp_fq -db $mirnaref_dir -o $out_dir/$sample

Miraligner stdout:

nov 18, 2018 12:12:13 PM miraligner.map readseq
INFORMACIÓN: Sun Nov 18 12:12:13 CET 2018
nov 18, 2018 12:12:15 PM miraligner.map readseq
INFORMACIÓN: **Number of reads to be mapped: 140399**
nov 18, 2018 12:12:15 PM miraligner.map readseq
INFORMACIÓN: Searching in precursors
nov 18, 2018 12:12:24 PM miraligner.map readseq
INFORMACIÓN: Sun Nov 18 12:12:24 CET 2018
nov 18, 2018 12:12:24 PM miraligner.map readseq
INFORMACIÓN: **Num reads annotated: 16674**

Output number of non mapped reads:
$grep "^>" ./miraligner_out/$sample.mirna.nomap | wc -l
**108194**

The number of initial reads and annotated are correct. From the output number of non mapped reads I conclude that there are 140399-108194-16674=15531 reads that it seems have been discarded, I don't know what's wrong with them. I would like to better understand how this tool works just to be sure that the counts I get from mature miRNA are correct to continue differential expression analysis.

Thanks in advance,
Ana

Reproduce the reported miraligner command with the collapsed-read input and compare the reported mapped, annotated, and .mirna.nomap counts. Trace how miraligner handles reads between the logged total and these output categories; done means explaining the 15,531-read discrepancy or identifying a reproducible counting defect.

Written by the indexing model from the issue text.