Most of the positions they point to are library artefacts. This tells you which ones are not.

conda install -c conda-forge -c bioconda hgt-site

hgt-site run --bam sample.bam --kraken-db /path/to/kraken2-db \
    --host-taxid 9606 --threads 8 -o sites.tsv

chrom      start    end      left_breakpoint  split_reads  n_pairs  top_taxid  qvalue
chr21_20M  1870963  1871693  1871390          4            16       562        1.295e-12
chr21_20M  2646531  2646869  2646659          4            5        562        3.33e-08

Every candidate stays in the table with the reason it failed, so you can see what was considered and rejected. Full options under Usage.

hgt-site detects candidate horizontal gene transfer and integration sites from short-read sequencing data. It finds places where DNA from bacteria, viruses or other organisms may have been inserted into a host genome, such as a human genome. Most bacterial reads in these samples come from contamination or from DNA fragments joined by chance in the lab, so hgt-site checks each possible site against what chance alone would produce and keeps only the sites with clearly more support. In the…