You can find some statistical power with any number of observations. Doesn't mean results can be generalized, which is what the article wants to do. 20 extremely low. At that level it is more reasonable to ask why anyone would assume there to be no differences between two groups of ten people.