We're working with a de novo transcriptome of a non-model organism. There were some genome assemblies for it, but we decided not to use it for GE analysis since they weren't annotated and also looked suspicious in some ways. Yet we've tried to map rRNA/mitRNA filtered reads on each of them and it appeared that 7-20% of reads wouldn't map on any variant of subject genome, so it's a contamination. We've mapped unaligned reads on genomes of every organism we're also working with, but there weren't any serious match (0-1%).
What could it be? Is it for sure contamination of some kind or might be something else?
Should we still work with all reads, or
choose those which are aligned on a genome and assemble a de novo transcripts with them?
UPD: The organism is basidiomycete fungus Lentinula edodes.