very long variants in VCF after giraffe alignment

Question

[Michal Frenkel]

I am utilizing VG for giraffe alignment to generate a VCF file (by running VG Giraffe, VG Pack, and VG Call). Upon inspecting the VCF, I occasionally encounter exceptionally lengthy variants spanning tens of thousands of bases. These variants exhibit a depth (DP) of 0 and possess low-quality scores. I am curious about the origins of these extensive variants. Furthermore, how does the tool handle the presence of heterozygous reads in the variant location, and what impact does this have on variant calling?