I think we all accept that a sim is only the deluxe version of MS FS with the $20m+ hardware upgrade .... and that it varies in fidelity according to how the software is tweaked.
If we play somewhere in routine-middle-of-the-envelope areas, then the fidelity tends to be good as it is subject to regular assessment and acceptance. Conversely, the more we move away, the more questionable might become the Type/model fidelity. This doesn't necessarily mean that the machine ceases to be useful ... it might just mean that the observations are generic rather than specific to the model alleged on the data plate and, at times, might be quite unrealistic ....
This was brought home to me some time ago when I was working with a particular sim which had a significant software upgrade affecting a particular primary control .. I had the opportunity to test the before and after versions with the installing software engineers (something about no-one else being interested in doing this at 3 am on a Sunday morning as I recall ..) and the differences in the envelope region affected were quite astounding ... the "before" characteristics were unrealistic .. the "after" very much like the reported flight article's behaviour.
I find it a little sad that so many people use the devices only for very routine, stereotypical training and proficiency checking .. the opportunity to give crews exposure to areas a little out of the ordinary (especially for building confidence in the more routine manipulative areas) is discounted ...
... and, of course ... we ARE trying to build skills which might prove to be of use for the flight article .. if one only wanted to play with the toy .. then the MS article is a lot cheaper ...