Yes, I pretty much agree with all that. I had an interesting conversation not long ago with Vittorio Gallese, who was one of the people who identified 'mirror neurons'. Gale's view has changed since in the sense of coming to think that 'mirroring' is a much more general property of neurocomputation and may in a sense be the default. The has an interesting implication for 'theory of mind' and autism that may suggest the usual theory is a bit backwards. It is clear that a lot of very basic perceptual functions depend on high level abstract inferences before even 'raw data' is defined. So we do not see the colour of an object until all sorts of compensatory tricks have been run that modify the colour according to what sort of object the brain thinks it is dealing with. The means that computations relating to moving an arm may well make use of an abstract concept of arms in general rather than specifying 'my arm'. So in a sense the mirror neuron is more basic than a neuron that recognises a movement as mine or hers. It is just a recogniser of an arm movement. If so there may be a case for saying that the autistic person is not so much incapable of recognising behaviour as belonging to someone else as incapable of recognising a movement as being of a generic type that is likely to have the same significance whether here or over there.