While I find the 17% number interesting to think about, "most" is far less interesting. Multiplication doesn't care about order so you're instantly cutting 2^64 possibilities down to about 2^63. That's a hair's breadth away from "most" already, and considering even a tiny amount of overlapping results gets you there.

What gets interesting is actually trying to quantify the overlapping results.

  {255 through 226 not used}   15 \* 15 = 225 
  {224 through 211 not used}  15 \* 14 = 14 \* 15 = 210 
  {197 through 209 not used}  15 \* 13 = 13 \* 15 = 195    
  14 \* 14 = 196
  {183 through 194 not used}   14 \* 13 = 13 \* 14 = 182

The chance of a random 64 bit integer being a 32 bit integer is 0.0000000233 %

The chance of a random 64 bit integer being a product of two 32 bit integers is 17%

Nice

[0]: https://en.wikipedia.org/wiki/Benford%27s_law