Re: This is not "mind-controlled".
Let me start off by saying you're right about the bit about the eyes, I misread that part of your post, my bad. Now to get to the meat of the question....
You seem to be proposing that the only acceptable control method should be a third-person visual image ("imaging what you want the robot to do and the robot doing it"). I don't see why a specific visual frame of reference should be the only acceptable method of control. Why not a first-person visual reference (Imagining oneself to be turning left, which is actually the "Avatar" method)? Why not an internally voiced audio command (internally thinking the words "turn left, you damn robot!"). And indeed, why not imagine oneself controlling a joystick that is in turn controlling the robot? These are all valid internal mental representations for the desired outcome.
Indeed, the way the mind works, it's highly likely that the mind will automatically be correlating different representations and firing them simultaneously. For example as someone who has used a joystick to play video games, my knowledge that the (real or imagined) joystick is controlling a robot will automatically correlate with a mental image of the robot turning left if I am using the joystick to direct it left. Quite possibly I'm at the same time issuing verbal instructions (surely I'm not the only one who has ever talked to their video game characters even though I know it won't have any effect).
In my book, as long as the computer is scanning mental patterns, identifying a recognisable pattern and correctly performing an action based on that pattern, the researchers can rightly call it mind control. It might be mind-control v 1.0, or maybe even v0.1, but still mind control.