yep! that's an option, we even have a tutorial for grabbing the 16S RefSeqs here:
https://forum.qiime2.org/t/using-rescript-to-compile-sequence-databases-and-taxonomy-classifiers-from-ncbi-genbank/15947/10
that might be a job for q2-clawback (to weight taxonomic classificaiton by the likelihood of detecting specific taxa, instead of trimming a database, which can lead to misclassification issues).
Good luck!