You’re not quoting the overlap, you’re quoting the (IMO) misleading framing that compares extremes. If you look at the picture, roughly one third of the distribution has overlap. It’s not a majority, but it’s certainly not negligible either.

BTW, the 0.1% number may be sampling noise, as similar studies have come up with numbers like 1% or 3%. While they’re all small, 3% is 30x larger than 0.1%. Don’t put a lot of stock in that number; it’s trying to draw conclusions from the noisiest lowest-data part of the distribution, and therefore is highly prone to error.