My guess: The Head Turner is one behavior and gets the camera feed. Under each turn there is a separate "Head". The Face behavior has to be bound to each of those heads correctly - it is the behavior that makes the head bigger and smaller. I am wondering if it only bound to one of the profiles (or none). To check in the Face behavior properties, there is a "Views" section and a "Handles" section. If you expand the Views section, did it find all the views OK? Then under the Handles section can you expand Head - did it find a head per view? Could you provide a screen shot of these two expanded perhaps?
I am wondering if it either failed to find the heads until each profile, or it bound to a handle in the layer rather than the layer itself. (I am a bit murky here, but I think if it says .../Head:Head then it bound to a handle inside the head layer, whereas if it says ..../Head it bound to the layer itself. I am not sure what Photoshop does, but in Illustrator I vaguely remember when it bound to a handle on the layer (rather than tagging the layer itself) it could not do as much. Try tagging the layer rather than putting a handle on the layer in that case. But I also think Photoshop was a bit different, so I am not 100% sure to be honest - maybe look at what the sample puppets that work do here. But that is where I would focussing.)