Statistics would say that if you have a representative sample of 10% of the population, the percent failed welds can be extrapolated to the population. So if 10% of the 10% are bad then you can be confident that something like 10% of all the welds are bad, rarely an acceptable outcome on a job where you are bothering to test.
Like the other posters, I would look at the data and any welder with a failed weld would get 20% X-Ray the next day. If ANY of those welds fail I would do 100% of that guy's welds until you get around to running him off.
David