>Simulation requires representation, which requires requires remaining differences unrepresented
Right. But presumably not all physical properties of brains are necessary for consciousness. For example, the fact that action potentials happen on the order of 1ms to 1 second probably isn't intrinsic to consciousness. That is, there is no in principle problem with having consciousness supervene on neural networks that have much longer cycles for the basic substrate of communication. It also probably doesn't matter that neurons communicate through opening and closing ion channels, are made of carbon atoms, are constructed by protein assemblies produced from DNA, etc.
What we need is to represent every necessary feature and relationship involved in manifesting consciousness. As a first pass estimation, it seems very likely that only the information-bearing structures are necessary for consciousness. But the information-bearing structures are substrate independent. It follows that a perfect simulation of these information-bearing structures engaged in their typical dynamics would be conscious.
>You can simulate the interaction [...] but the two are not the same.
This idea that something needs to be "the same" is doing a lot of work in your argument. But most relationships involved in brains are incidental to consciousness. We need to get clear on what the necessary features of consciousness are. This is the only measure of sameness that matters, not superficial resemblance.