Why is VCFTools excluding so many SNPs in its Tajima D calculation?

Question

[Ace]

I used VCFTools to calculate Tajima's D for my population. When I started doing more specific analyses, I found that while VCFTools (via allele freq) recognized 22 SNPs in a certain gene, it was only including 5, total, into all the regions that fell into that area, producing many NA windows. Is this an outcome of certain parameters required for inclusion into its TajimaD testing?

[Mark]

Ive got no answer apart from try to use vcf-kit instead.