I have not heard of this being done (so check the lit before listening to me!), but just want to give an idea of how to test this question:
Use RESCRIPt to create a plastid 16S gene database, and use the same plugin to test taxonomic resolution. That would validate at which level you can reliably distinguish plants (family? species? strain?), so maybe enough to justify this approach if you are looking at broad phylogenetic groups.
But first you would also need to validate that you have adequate coverage of plastid 16S seqs using your same PCR primers...
good luck!