Hi @ebolyen, I also noted the exceptionally high number of unique features before removing chimeras. I reran the analyses and found that unique features were reduced from 7590 to 2535 during the chimera removing. However, I didn't see a significant drop in sequence number per sample (5% on average).
dim(seqtab)
[1] 80 7590
dim(seqtab.nochim)
[1] 80 2535
For the amplicon sizes, I did have the same doubt. The observed amplicon sizes are within the expected size ranges based on the in silico evaluation.