I think using an aimoffset would be better than trying to control the individual bone. Aimoffset is used in the animation content example to make the controllable pawn aim where the camera is looking, but you could make reference animations with just the head moving instead of the entire body and then use a layered blend per bone with the neck as the beginning of the blend.
Use interpolation to smooth out movements, the node is called rinterp to. Get delta time from your event update animation, read the current value of the rotation, the new calculated value of the rotation should be the target and then set the return value as the current value. The example below is for getting where the third person camera is looking and limiting the angle to between -90 and 90 so that it can be used with an aimoffset asset.