What you are asking for is not feasible. Any method used will be imperfect since it can't uniquely assign every read (let alone bases) that belongs to your organism of choice, especially when it is present in a metagenome.

One option is to try mash and RefSeq or just your genome. You could also use bbsplit.sh from BBMap to bin your reads so the ones that best map to your genome can be separated. bbsplit gives you multiple options to handle reads that multi-map within and across references genomes.