First - I am aware of all that, though I fail to understand how rotational momentum works against translational movement. It seems EXTREMELY counterintuitive. As far as I know, gyroscopes only stop rotational movement in the plane they're not spinning in (which, in my point of view, is absolutely perfect for a bipedal considering the position of the hips won't vary much more than five degrees - unless you want to design a crawling robot, or something, it is not really an issue). Place the flywheel "horizontally" and the robot can spin around itself without falling. As for motorcycles, it is a completely different example - flywheels, in that case, are not adequate because a motorcycle's agility depends on two planes of rotation - yaw and roll. But seriously though - if you have proof that gyroscopes stop translational movement in it's frame of reference, let me know - because as far as I know, the ISS's not showing any signs of falling down, and it's not geosynchronous.

And as @acyclicks mentioned, if your robot needs to be angled at more angles than would be allowed otherwise, you can mount it on a rig with actuators or within a sphere with motors to change the angle.