I am speculating here.

The reason for it being a two level device at present is likely due to it being mostly research and not so much engineering.

They say their next chip will deliver grey scale and many more signal points.

My guess on color is one or more of the following is true:

[0]The color info is normally sent via the color sensitive cells now damaged and we have yet to understand how that signal enters the nerves we can send a signal to.

[1]It may be that we need a far smaller, more precise signal point to achieve color. Current tech stimulates many nerve endings. This was the basis for my interpolation comment above. Basically, each pixel stimulates an area of the damaged retina which contains a great many possible signal points if it were possible to stimulate them individually. Because so many are stimulated all at once, the subject perceives white phosphines rather than colored ones.

An analogy would be the colors on a CRT. A broadband beam would light them all up, yielding monochrome vision. A narrow beam can light up a few or just one, yielding color.

One thing I just realized writing this is our blue sensor cells are scattered about, not well clustered like the green and red ones are.

Maybe current users see a bit of color at the very extent of the artificial visual field due to a failure to hit the necessary blue cells...

[2] It may be some sort of pulse is needed to encode colors. And perhaps the current signaling is continuous.

Hopefully, we get an answer from the team.