It's actually L515 and L516 (20/17/15) that have a number of turns lower than their 'design' inductance.
L517 and L518 are right on target.
You can see this by using toroid.info to calculate the number of turns for a given inductance and toroid type.
I suspect that the reason L515 and L516 have a low number of turns, is that during testing Hans found that the lower numbers worked better than the original design simulation.? He does clearly state in the assembly guide to not try to measure/match the inductances, but to use the number of turns given, then change from there as needed.
