MetaBat2 contig binning question

[jmartin]

I have a collection of metagenomic samples and I want to look at genomic microdiversity between these samples using the 'inStrain' tool. To do that I need to build a genomes db, and the documentation recommends to do de novo MGS assemblies using data from my samples to ensure that my genomes db has the specific genomes that exist in my samples (as opposed to just the closest genomes found in the public repositories).

So I've assembled each sample (using metaSPAdes & MEGAHIT), merged the resulting contigs into a single db and mapped each of my sample's reads against that db, and then I fed these many bamfiles to MetaBat2, which produced ~8k genome bins, which seems reasonable for my data.

But I'm now having a problem understanding how this should work. And its probably just my own lack of understanding of contig binning that I am hoping people here can help me with. I feel like each of my genome bins generated by MetaBat2 may have overlapping contig data. It makes sense to me that the contigs are correctly binned by genome, using the depth information per sample and similarity. But from what I've read I don't see anywhere that says each genome bin has been 'flattened' down to just the consensus of the assembly contigs.

So my question is: After doing contig binning using MetaBat2, do I need to build a single consensus per genome bin? Or has that already been done? Or do people even worry about having overlapping contigs in these genome bins?