Contrary to popular belief, the position of the arm related to the servo has
absolutely no effect on control surface differential. the only parts of
the servo arm setup that has to be on a 90degree angle is the pushrod to
the arm. A servo travels the same distance no matter what angle it is
installed, but if the pushrod to arm angle is off it will induce a
differential of the control surface deflection. On most of the planes
we see the pushrods are parallel to the servo case, so CGretired's sketch
is very accurate. but I have played with small electrics and a few poorly
built planes that the pushrods were set at wierd angles to the servo
and on a deflection meter you could see that a differential was present.
this was easily fixed by setting the servo horn at a 90degree angle to the
pushrod. I made a slight modification to CGr's sketch to show what I
mean (the blue lines represent the pushrod and the red lines show the angle
between the pushrod and the servo horn).

This same principal applies to the control horn at the control surface.
an optimal setup is a 90degree connection of the pushrod and control horn
continuing down through the hinge line. I have added a picture (I am no artist)