That is a spectacular image. Thanks for showing.
Here's what I'm finding - although the laser may well be modulated - and although the response be fairly linear ratio of optical power out for amps in, the alternative of burn/not.burn is essentially binary. So it seems that the 'grayscale' is being produced by widening of the lines that varies with the joules/second delivered to the work.
For a line to widen there are only 2 ways I can imagine- combustion, i.e., material burns that is not directly heated by the laser beam, or else the beam is wider than the effective burning 'circle' in the center of it and that expands with intensity. Can you confirm or correct that?
Regardless of the exact mechanism, though, here's what the results show:
http://imgur.com/a/dLH9o
As can be seen, the best burn (10 bit modulation) is replicable with no more than 16 shades.
In a lame analogy, the speedometer may say 1000 MPH, but the car isn't goin that fast no matter what.
In fact, the color depth can be reduced to about 9 shades before there's any noticeable degradation of the image.
At the moment, my little laser is on a cnc mill, so the speed is quite slow or i'd be burning a copy of that picture for a test. Perhaps I will make a machine optimized for lasering so I may dare to attempt the larger images... or- would you be willing to attempt burning that if I provide gcode for a true halftone?
Or maybe I can just do a part of the picture for a test... lemme work on that...