Such tests are surrounded in controversy. These ancestral percentages are easily confused with cultural definitions of race and ethnicity. (For example, "African ancestry" does not equal "black.") The accuracy of the tests is also debatable. Some critics charge that the sample populations used to determine which SNPs are associated with which ancestries are too narrow. Sequencing technology continues to improve, but even a small percentage of mistakes can lead to a large number of misread bases when scanning tens of billions of nucleotides. (A genome must be read several times for accuracy.) The company that did the sequencing claims that each base was read an average of 7.4 times, but Kari Stefansson, whose company assessed Watson's heritage, says he found enough errors in the public genome to have doubts about whether the 16 percent figure will hold up. For example, he says there are places where it appears that Watson has two X chromosomes, which would make him a woman. ["What does it mean to say James Watson is 16 percent African?"]
Stefansson is basically admitting he will trade scientific credibility for publicity and PC status points. (Link via a comment by "larch" at MR).