Almost Nothing Mapped Using Bwa Or Bowtie. But A Lot Of Mapping In Blast

I have pair-ended illumina samples with the read length of 100-150bp. I tried to map them to the reference genome/transcritpome, but almost nothing mapped using BWA or Bowtie.

When I blast them to nt reference using megablast, the most reads map to the right species. What does this mean? How can I interpret this? and what do I do to improve short-read alignment using BWA or bowtie?