When targetting at the bones, the aimbot targets only one pixel of it (even though it can be moved up or down with extra height aim), making it look like a perfect central snap, which it can be detected as aimbot.
So I thought, the aimbot target would be separated, and the aim points would randomize it's X and Y as the aimbot turns on, so it looks like a normal human pointing and normal human mouse movement.
Here's an example of how multi-pointing would work:

On the top part (single-point), the black circle represents the head of the model, The green box represents the head hitbox, and the cross represents the aimpoint. The image shows that the aimbot only aims central, where the aim point is shown, even with smoothing, it looks like a perfect direct movement to the center of the head.
On the bottom part (multi-point), it shows many crosses, which those are results of each time the aimpoint's location randomizes, making it look like there's no determined central pixel for the player to aim at, making it look like human targetting, the crosses at the edges of the hitbox would be helpful when an opponent peaks out, and half of his head is invisible.
(Non-important part)This would work on different hitboxes on the model differently.