How to deal with sequencing reads that can both align to miRNA precursors and mRNA database?

Question

[dolphing]

Hi,

I have some miRNA sequencing datasets from human, the 3' adapters were trimmed and reads length less than 18 bp were discarded. When I aligned these datasets to miRNA precursor database and mRNA database, I found that more than 80% reads can both align to these two databases.

If I want to predict novel miRNA, should I remove the reads aligned to these two databases or only remove those aligned to miRNA precursors?

Thanks in advance for your help!