Low mapping rate during Bowtie

[SaelinB]

I am trying to use shortstack for de novo miRNA discovery, and have little experience with small RNA. When I run the software, only ~33% of my reads map:

Uniquely mapped (U) reads: 7153136/76751715 (9.3%)
Multi-mapped reads placed (P) with guidance: 14066544/76751715 (18.3%)
Multi-mapped reads randomly (R) placed: 3615495/76751715 (4.7%)
Very highly (H) multi-mapped reads (>=50 hits): 598454/76751715 (0.8%)
Not mapped (N) reads (no hits): 51318086/76751715 (66.9%)

I also only get 1 miRNA predicted in the final output

I pre-processed my reads with trimmomatic, removing adaptors and keeping reads >= 15 nt: java -jar ~/opt/Trimmomatic-0.39/trimmomatic-0.39.jar PE -threads 12 -phred33 ${i}_combined_R1.fastq.gz ${i}_combined_R2.fastq.gz ${i}_1_P.q25.fq ${i}_1_UP.q25.fq ${i}_2_P.q25.fq ${i}_2_UP.q25.fq ILLUMINACLIP:small_rna_adaptors.fa:2:15:8:2:True LEADING:3 TRAILING:3 SLIDINGWINDOW:4:25 MINLEN:15

This is paired-end data, but I then only used R1 and unpaired R1/R2 reads from trimming. I also removed reads > 60 bp. FastQC length distribution shows a large peak at 17bp, and a smaller bump at 24bp.

Do you know of anything that would lead to a low mapping rate, or is my data just bad?