I do see what you mean, it's beyond me too, but it doesn't matter that you don't understand how it works, just as long as the people using it do.
A lot of words seem to get used incorrectly.
Alleles is a word we should be using. At each site a person will have 2 alleles (one from the mother and one from the father). It is possible that at that site the allele inherited from the mother will be the same as the father so they will be homologous at that site.
So if the lab checks 10 sites there could be a maximum of 20 alleles. If they found 3 alleles at any one site it suggests at least 2 persons contributed to the DNA sample, but it could be 3 persons if they were all contributors were homologous at that site.
If they found 5 alleles at any one site it suggests at least 3 persons contributed to the DNA sample, but it could be 5 persons if they were all homologous at that site. (This is the situation found in the MM case IMO)
But if there were 3 persons contributing to a DNA sample there could be a maximum total of 60 alleles in the sample (6 alleles per site).
If they only found a total of 36 alleles from a sample that had at least 3 contributors IMO it suggests the contributors were closely related.
If the sample was a mix DNA from child, mother and maternal grandmother, I think it is possible to have a total of 36 alleles, but from analysis there would also need to be multiple homologous alleles between the father's and the maternal line. This could happen in smallish communities.