Says this guy. While the author seems to be motivated by race-denialism, I think he's probably right that "forcing STRUCTURE to place individuals into too few clusters" can lead to non-optimal results (which would be consistent with what Dienekes appears to be finding). Certainly, contra Rienzi's railing against "incompetent amateurs, with their K=15 ADMIXTURE plots and clustering with up to 124 components", there's no reason to assume that in general runs with lower K will give more meaningful or accurate results than runs with higher K using programs like STRUCTURE and admixture.
I'm not terribly interested in Kalinowski's specific attempt to prove that African farmers are genetically closer to Europeans than to African hunter gatherers, but two things: (1) if this is true, I'd say it's much more likely to be attributable to back-migrations from Eurasia than to Europeans being descended from "African farmers"; (2) I enjoyed the section of the article in which the author promotes neighbor-joining trees as a superior method for visualizing relationships between populations, only to go on to explain that "even a tree with a R2 of 0.98 does not accurately depict all of the relationships between populations" since it still "depicts all sub-Saharan African populations as being more similar to each other than to European populations."
Reference: KALINOWSKI ST (2010) The computer program STRUCTURE does not reliably identify the main genetic clusters within species: simulations and implications for human population structure. Heredity (Published online). pdf