I would not be surprised if a factor was mixed in for the "most rapes go unreported, especially date rapes," estimates.[1] Given these figures, either the 3% conviction rate or the 46% reporting rate, a person could take numbers that may or may not already include these adjustments and apply the most helpful. 10,000 x 33 = 330,000, might as well round up to 400k.[2]

I'm not saying that's what happened in this case, but abuse of confidence intervals is an common problem, even among those with good intentions and little idea that what they're doing is misleading.

[1] http://www.rainn.org/get-information/statistics/reporting-ra...

[2] This guy says 400k, but we know that less than half are ever reported, so it's probably closer to a million. And so on.