That's not quite correct. The calculations disproved it to a certainty of one against a million. Or, to put it in another way, the data they had clearly showed

That's not quite correct. The calculations disproved it to a certainty of one against a million. Or, to put it in another way, the data they had clearly showed that it couldn't happen, but taking into account all known sources of errors, there was a one-in-a-million chance that the data they had was the result of random errors.