I can't help but notice your table only goes to 511. That's half of the available pwm pulse width.
that's correct dave.
The point was to demonstrate that its possible to eliminate the visible brightness steps that are most evident at the dimmest fading levels. The steps are quite visible at the lowest 20 or so pulse widths wether its gamma corrected at full 10 bit resolution or linear at 8 bit resolution .
100 steps was all that's necessary to prove the point , it can be done

the missing 10 steps
100 481
101 522
102 568
103 617
104 671
105 729
106 793
107 863
108 940
109 1023