I have some 16S data, and encountered a problem while trying to build an insertion tree using the qiime fragment-insertion sepp function. Although I know that I have ~18000 unique ASVs, when I tried to build a tree from my representative sequences (filtered to the ASV level and trained on silva-132), I get an insertion tree that only has 196 taxa. My initial thought was that maybe my sequences are just incredibly different from what is found in the silva database-- I know the sepp step gets rid of sequences that are too divergent. But I compared this to a greengenes insertion tree generated by Qiita, and it only had 193 taxa. I’m confident that my sequences are not so novel that over 17,000 of them are less than 70% similar to the databases.