I'm having a hard time conceptualizing why changing the stringency option when trimming adapters is increasing the number of discordantly aligned reads during mapping (bowtie2). When increasing the stringency from 1 (the trimgalore default) to 3 (cutadapt default), the number of discordant reads goes from 58,336 -> 513,581. From how I understand it, higher stringency should lead to fewer false positives when adapter trimming at the cost of potentially missing false negatives in those 1-2 bases... but why would this lead to more discordant alignments? The experiment is an ATAC-seq experiment with 75bp reads. Please let me know if you have any ideas or know what is going on here!