Thanks for reaching out and for providing those sequencing run screenshots! Here are a few thoughts on how to proceed here:
As of the QIIME 2 2021.4 release, DADA2 requires at least 12nt overlap for proper merging - so that's something to keep in mind if you only have ~100 bp of overlap.
I do agree with you that your forward reads are pretty low in quality, and may not be good enough to use at all (it appears as though the only semi-decent quality score is from 0-10 in your sequence base, which is pretty short).
Using only reverse reads in denoise-single isn't typically recommended due to the commonly lower quality scores of reverse reads, but in this case it does seem like your best bet. In this forum post, this is what @Mehrbod_Estaki recommended to another user who has poor quality forward reads as well. Hopefully this helps!
Cheers,
Liz