Please provide more information. You've shown a command to trim PE data, but haven't said what you're trying to achieve. But yes, trimming raw reads before alignment is generally a good idea, but it depends on what the QC looks like, which you've also not provided.
1) if the library was done properly, it's very unlikely you have many fragments shorter than your read length. You probably won't see much adapter at all.
2) splice-aware aligners don't mind soft trimming to make an alignment. Even if you had adapters on a read, it will probably still align to the right place, and be counted as belonging to the correct gene.
Now, if you were doing DNASeq, or do novo assembly, then removing adapters would be more important. But your typical RNASeq experiment, where all you want are gene/transcript counts? Don't agonize. It's not going to be much a problem, even if you did nothing.
Please provide more information. You've shown a command to trim PE data, but haven't said what you're trying to achieve. But yes, trimming raw reads before alignment is generally a good idea, but it depends on what the QC looks like, which you've also not provided.