Different genotype of SV for the same sample in different batch

[lemon-latte]

I am using smoove for SV calling for population-level calling, Using smoove merge to get the union of sites across all samples, and for the genotype step, I used the different strategs 1) genotype each sample and paste all the single sample VCFs togather; 2) extract some samples to genotype and paste all the single sample VCFs togather. And I find there are some differences in genotypes for SV with different batches. I don't know what results in the different genotypes.