Why do different users have different number of SNPs on the same 23andMe genotyping chip?

Question

[mprza]

When looking at two individuals 23andMe outputs, I notice that they have a different number of SNPs. Each file records homozygous variants, heterozygous variants and uncalled variants. So, why would this be the case? If 23andMe looks for the same set of selected variants, shouldn't the sum of homo, hetero and uncalled variants be equal for every person?

I understand that some people just might have more variants than others, but surely the variants which they don't have would show up as uncalled or homozygous (to reference allele)? I say this because I can see other homozygous (reference allele) and uncalled variants in the 23andMe data.

[GenoMax]

So, why would this be the case?

Quality control may be resulting in filtering out of some SNP's.

[mprza]

There are quality checks in 23andMe, but this page indicates that genotype calls which are not confident should still be reported in the data, but labelled with a "--" (https://customercare.23andme.com/hc/en-us/articles/212196888-What-Does-Not-Determined-Mean-)