This kind of thing always has the same weakness: it's extrapolating from a sample size of one. It of course tries to account for that by adding large error margins based on what we know, but for some of them it's pure guesswork that could easily wrong.