I think a stronger bolt spring is all it needs.

What I think is happening is this:

When the gun recoils, the muzzle jumps up. The inertia of the bolt makes it want to pull out of the notch in the cylinder. At that point the cylinder is free to rotate either direction.

On shot #1 out of a full cylinder, there is nothing to make the cylinder want to rotate. But on shots #2 thru 6, there are empty cylinders on the left side of the gun. When the muzzle jumps up, the inertia of the live ammo on the RH side of the cylinder puts a clockwise (as viewed from rear) torque on the cylinder. If the bolt simultaneously gets pulled out of the notch, the cylinder rotates backwards.

Should be an easy fix. Stronger bolt spring.

Have you ever noticed it doing this with shot #1? If so, my theory may be blown out of the water.

edit: I also think that a ported gun would not have this problem as the muzzle jump is reduced quite a bit.