We have paired end Illumina HiSeq 4000 reads that we are working to remove adaptors and trim bases below our phred cutoff. Trimmomatic has 4 output files from our two input files. Two that are paired and then two that are unpaired. I was originally going to toss the unpaired trimmed files and just run trinity de novo assembly using the paired trimmed files but my co-advisor mentioned that we may want to use both. He is saying that trimmomatic when cutting the adapter sequences, may also cut the complementary sequence to that adaptor on the paired strand leading to removal of sequences that we don't want removed. I would assume that trimmomatic would only identify the actual adapter sequence within the designated forward or reverse read but now I am having trouble finding the answer to whether this is the case or not. He mentioned using both the paired and unpaired reads in the trinity de novo assemble, that we may be able to recover erroneously trimmed segments.
What are people's opinions/experiences on this matter?