I actually have to revise what I had stated earlier. Clamping one of the values to 255 is not going to work. The way the rings are generated in my lighting system is that the optics in fromt of the LED head " clamps" one of the colors to zero.
So in essence if your algorithm chooses one of the three channels to clamp to zero and then you choose values randomly for the other two channels it'll work better.
However, the approach fungus has posted, selectin HSV and conveting it to RGB - while more complicated to implement and more computationally expensive - is by far the best approach!