Potato has about 100 related wild Solanum species growing naturally in the Americas. The US Potato Genebank aims to keep samples useful for research and breeding to improve the crop, often in the form of botanical seed families. A key component of genebank efficiency is assessing diversity within and among populations, and DNA marker sequence diversity is a powerful proxy for trait diversity. We previously reported on three factors which can cause under-estimation of heterozygosity: ascertainment, allele frequency, and ploidy bias. We here report, using GBS data for four diploid potato species, that average percent of apparent heterozygosity increases as data is more complete—the maximum difference was 2% heterozygotes when only a few individuals are called, to 36% when nearly all individuals were called. However, there was evidence that estimates of average heterozygosity based only on loci for which every individual has data can also be biased upward. Implausibly high levels of heterozygosity suggest non-segregating non-homologous SNPs, which occurred as 5–9% of all loci with complete data. We propose that best estimates of average heterozygosity in unselected seedlings should be based on loci with data for all samples after eliminating those loci that appear to be artificially fixed as heterozygous, which reduces observed heterozygote frequency by 16–26%. On that basis, the wild species examined have similar heterozygosity to the cultivated phureja.