Reading between the lines of what is becoming a very interesting thread, I think the problem may lie in the way that sims are procured. There is obviously the impossible (sustained G motion effect) and the imponderable (sim performance outside the explored), but these only accounts for a very small number of niggles about sims.
In my experience, the situation usualy involves knowledgeable Aircrew who know what they want, and engineers who for the most part, are capable people who take a pride in getting it right. Ideal situation - what can go wrong? Well almost invariably there are at least two layers of bean counters involved in the equation.
I have always found that the shorter the line of communication I have with the end user, the better the job turns out.
Another problem is is the shear magnitude of the task of laying down the exact specifics of every single function of anything as complex as a simulator. May be a solution would be to have 'bedding in' period post acceptance. I think to be fare to both sides, this would have to be chargeable to the customer, but any design change/update required during that time would be at nominal cost rather than at full market premium.