I am using miRDeep2 to detect novel miRNAs in a species that miRNAs have not been studied in before. The program itself is running fine, but I am getting inconsistent results between my different samples (all replicates of each other, just different libraries). For example, in library 1, miRDeep2 might predict 9 miRNAs (the genome is very small so I don't expect too too many). But then when I run the script on library 2, maybe only 6 are identified, and of those maybe only 3 were also predicted from library 1. The weird thing is, when I take a miRNA only detected in one library, I can find reads (like thousands of them) exactly matching that sequence in the other library where it wasn't predicted. So my questions is, why would a miRNA be predicted from one library and not the other when reads corresponding to its sequence are abundantly found (at roughly equal counts) in both libraries??

Hopefully this makes sense. I am using default miRDeep2 parameters - I didn't feel providing the code would help here as its more of a conceptual question.

Additionally, when I run miRDeep2 analysis on both libraries together (by providing a config.txt file), I get even LESS miRNAs predicted overall than I would if predicting from one library alone. Why might this be?

Thanks for your help in advance! I'm very new to bioinformatics so any help is appreciated.