[biot]

  > Apple's implementation is also backwards. [...]
  > You'll notice that when changing the roll of the knob,
  > the shines rotate in the same direction (against the roll)

It's dependent upon whether the source of light is on the same or opposite side of the surface normal as you are. Try it again with the light source behind you and you will see Apple's implementation.

[mdznr]

You're right on whether the source of light is on the same or opposite side. I'm assuming the light is coming from above you and with respect to the view, in front of it. http://mdznr.com/bJ The way, I think, you're talking about is if the light were to be coming from the user's chest (certainly not what's expected and not in the HIG).

The part that's backwards about Apple's implementation is that tilting the device (roll) should rotate the shines and tilting the device (pitch) should change the size of the shines. Apple's rotates on pitch and changes size on roll.

[biot]

Coming from the user's chest isn't expected, but imagine a reading chair with the lamp over the user's shoulder, behind the chair. Same direction and a common scenario that I can imagine reproducing in software. How often do people place floor lamps directly in front of their chair?

If Apple's HIG says that the light source should always be directly ahead of you, then you are certainly correct. However, if straight ahead is 0 degrees then you can reproduce the opposite behavior (the effects of pitch and roll get swapped) by having the light source at 90 degrees (reading light beside the chair's right arm). The direction of shine rotation on the knob is reversed by having the light source 180 degrees from its current position (reading light beside the chair's left arm).