Not to risk my position in Darwin Central, but I think he meant three sets of twins out of a million people.
That does seem reasonable, but not relevant. Darwin Central encourages honesty.
I considered that, except for the fact that that would have made the (already flawed) analogy even *more* of a complete departure from a useful comparison to the SCID-X1 study results.
Even though it would still be a poor analogy, one could try to use the analogy of "hitting the same person three times", or "hitting each of a pre-chosen set of three people" or "hitting all members of a set of triplets". (It would be a poor analogy because the SCID-X1 trials actually hit the same region in three *different* blasts of the "machine gun", but at least the "three hits" notion is preserved).
The "three sets of twins", on the other hand, has *NO* meaningful equivalent in the SCID-X1 trials whatsoever -- there's no "triple pairwise" event.