I am currently working on an “ideal brain” model (think of physical science's “ideal gas” model with neurons instead of molecules and links between neurons like collisions between molecules). One of the features of this model is that any network of neurons can work in two ways – recognising or doing – and the two roles are dynamically interchangeable. This suggests that there is not a class of “mirror neurons” because in theory all neurons can work in this dual manner.
Of course some activities are more relevant to recognition and some more relevant to doing, and there are some where it would be difficult to carry out the kinds of experiments that lead to scientist postulating that there was a special class of neuron.
What is interesting is that a basic feature of how neurons work could also be important in understanding how animals and people interact. The relevance of the mechanisms to empathy and social interactions is extremely interesting – but one must be careful to recognise that how much attention the mind gives to something will be affected by how relevant the activity is seen to be – and the observed levels of “mirror neuron” activities in experiments may be more related to motivation to pay attention than any specific factor in the part of the neural network being monitored.
What I find more interesting is that the ability of the mind to mirror what other people/animals are doing could be very relevant to some kinds of learning. Our brain sets up a neural pattern which recognises the activity and the tries to execute it to repeat the actions.