Hi @Will_Rumfelt,
Just to follow up...
If you'd like to reduce any potential negative affects of incomplete or mis-annotated taxonomy, you can remove reference sequences that do not have at least a family or genus-level taxonomy. You can achieve this by using taxonomy-based filtering of the sequences. Then proceed with any other filtering that you require prior to finalizing and training your reference database.
For any retained sequences that you know are mis-annotated, you can use RESCRIPt's edit-taxonomy action to fix these.
-Cheers!