Whether brains "are" TM's may be mostly moot from a simulation/emulation/modelling standpoint. We may not need a perfect model to get a "good enough" model. Biological systems are generally designed to handle a degree of noise and damage and still function fairly well. Thus noise introduced by the imperfections of a model may similarly not sufficiently degrade its ability to act like a normal brain.
On a really small scale our models may lack certain properties of biological brains, but it may not matter for most emulation. How much detail and/or actual "mirroring" is necessary to get good-enough emulation is still an unknown but key question.
And, I suspect we may find that fully mirroring biological systems is not the most efficient way to design good AI similar to how flapping wings were not the best way to get powered flight.