I'd like to see a reference for that statistic.

I don't discredit the statistic, but I think it's bit misleading, because we're not able to dig into the data - even just a bit.