I successfully got the pre-made Silva 138 classifier to run and tested it against my samples.
One thing I did notice was that when I used the qiime taxa-filter table command to remove chloroplast and mitochondria sequences, which is what I included in my original code with data from the 132 database with Silva, is that a hit for Eukaryotes shows up, specifically for Chlorophytes, or a species of the Desmodesmus genus. This appears in one of my field samples in the same level of abundance using the 132 database was classified as "ambiguous taxa", making up ~2% of the relative abundance of whatever else was present
Not sure if this was incorrectly characterized, or why this would appear after removing chloroplast sequences...
That is, Bacteria, Archaea, and Eukaryota are present. Eukaryotes, can be detected either by the mitochondrial / chloroplast 16S rRNA gene and/or the 18S rRNA gene.
So, even though you've removed plastid sequences (i.e. mitochondria and chloroplast), any 18S off-targets may remain. It is quite common for Bacterial & Archaeal specific primers to amplify eukaryote off-target sequences. Simply add "Eukaryota" to your exclusion list if you are only interested in Bacteria and Archaea like so:
Also, keep in mind that the QIIME formatted SILVA files provided by QIIME 2020.2 and earlier, differed in their preparation. Additionally, taxonomic and sequence curation can change between SILVA versions. You can read the original README file contained within the Silva_132_release.zip located here. However, you can use RESCRIPt to download and format fresh copy of SILVA 132 that is inline with similar processing of later releases. You can run the following to obtain the data and process how you'd like with RESCRIPt: