We wrote a pipeline for doing this here: https://github.com/MHH-RCUG/nf_wochenende

Check the installation docs. You can download a reference sequence of many bacteria which should include your samples of interest.

When doing metagenomics you are not working with isolates, but obviously with mixed samples, so I would always try to work with as many genomes in your reference sequence as possible. The flipside is you get problems with read attributions, mapping quality and quantification if you include say 1000 almost identical E. coli genomes in your reference. Try out our carefully constructed reference and see how you get on.

Suggest BBTools and minimap2.

With BBTools you can also provide multiple references with the bbsplit which is more for dealing with a mixed sample but might work for your use case.

Wth minimap2 you can set high stringency with the -x asm5 (most stringent) or -x asm20 (least stringent of these settings)

asm5/asm10/asm20 - asm-to-ref mapping, for ~0.1/1/5% sequence divergence