Hi @Francisco,
From a trimming perspective, I think the dada2 supports the shorter reads. (You're merging like 75%, wow!) You're losing most of your reads after the merge, in the chimera detection step. So, are you working with a COI or something where standard chimera detection won't work? Do you just have a lot of singletons because you have few samples and a lots of depth and therefore rare things are identified as chimeric based on the distribution? (This is a guess, I work less with DADA2 and Chimera removal). You could try doing DADA2 without chimera removal and then running through a secondary program to handle them.
Best,
Justine