From looking at the pictures of the grids provided on that website, I think I understand what they are doing. It's not actually very complicated.
It seems easiest to explain this if we consider only one of the hexagonal "tubes" formed by the grid. Normally, the light from a flash spreads out over some (relatively wide) angle, but now think of what happens to the light when it is confined to this tube: The light that would normally head off at a steep angle is intercepted by the walls of the tube, while light that was already propagating parallel (or at least close to parallel) to the tube axis is passed without obstruction. This results in a more confined range of angles over which the light can spread. Of course, the confinement is not perfect, as I will explain below, but you can imagine how the angular spread of the beam would be limited to consist only of those rays which can pass cleanly through the tube. This is simple geometry, and you could adjust this angle by varying the length and diameter of the tube.
However, the light that is intercepted by the tube walls is not completely eliminated. Instead it scatters into a random direction. It is this scattering that gives the resulting beam a nice soft edge. Instead of being sharply confined to the angular range allowed by the tube, the scattering allows it to "bleed" a little bit.
I should add that the hexagonal shape of the tubes probably has very little effect on the resulting illumination pattern. More likely a hexagonal shape was chosen for some other reason, like easy manufacturing or more mechanical stability. Optically, a square grid or even an array of circular tubes, like a bundle of straws, would work just fine.