The annual increase in number of dereplicated species representative genomes in GTDB or GlobDB shows how undersampled. And reference based MG profiling works best at species level, but not so great at higher phylogenetic distance.
Anyways, what I want to say is amplicon sequencing ain't dead (yet)