I had a quick look at the contigs that encode the three S. aureus 16S sequences (my own annotation, so not directly applicable to GTDB locus names): If I'm not mistaken they're either only a 16S or some rRNAs together with some tRNAs. Using other markers when present on the same contigs as 16S might hence be an alternative to only using the "contaminated" content in the NCBI file, i.e. if there are other markers -- preferably CDS, perhaps even one of the 53/120 -- disregard "contaminated". This is, however, what we wanted to avoid with the Sativa approach: If MAG assembly gets confused and actually assembles the wrong rRNAs into other contigs we wouldn't notice. I admit that the risk for this is smaller than binning causing erroneous contigs getting into a bin.