For ACS Ancestry data, 1st response (2007-2011, ACS 5-year sample). I compared the summarized microdata from your site to the summarys generated by Factfinder at the US census site. I expected them to be the same, but they are not. I know groups with under 10,000 people are rolled up for privacy issues, but is that the only thing that causes these variances?
example for the whole US:
ANCESTR1D IPUMS Census summary
Somalian 105,927 102,503
Kenyan 41,960 45,089
Its also not clear exactly what Ancestry group the omited ones are rolled up to.