Basically the FSX flight model is so fundamentally broken, it doesn't even understand the concept of gravity properly.
First of all there are the subtle handling qualities. Because the flight dynamics are in general totally fabricated (at least, the reactions they bothered to try and simulate) the whole aircraft just doesn't feel right. The most obvious example is prop torque. Non-existent in most FS aircraft, and where they did try, it is not nearly pronounced enough (the only people to really hit it are A2A Simulations with their range of WW2 and vintage aircraft). A friend of mine owns a Mooney - he says the FS flight model is totally dead compared to his actual aircraft. At takeoff he requires full rudder to counter yaw induced by the propeller (and that isn't enough until he gets to VMCG). This for example, is completely wrong in FS.
The stall in FS is screwed. Aircraft simply do not stall properly. Only one aircraft I know of (freeware) spins anything like realistically, but even that doesn't stall properly.
Angle of attack is totally broken. I only know this as I was a dev on a major project. We spent months trying to figure out why our FBW model wouldn't react correctly in high altitude flight regime, until we discovered AoA is CONSTANT regardless of altitude. In short: if the aircraft stalls at 100 kts at sea level, it will stall at 100 kts at FL400. COMPLETELY INCORRECT. As a consequence, we couldn't model other stuff correctly either. Forget trying to model stall characteristics in anything but clean configuration at sea level.
Take any jet aircraft you like to altitude and pull it around. Nothing bad will happen. Deep stall can't be modelled correctly either.
Inverted flight doesn't work properly either, and is related to the fact AoA is totally faked, and wrong at that.
All this on top of the fact that a shutdown engine has far too much drag, and can't be altered. Trying to correctly simulate engine-out is impossible, and is in part related to rudder behavior - the whole concept of yaw is screwed. Needless to say glide ratio is totally #%&*$# as well.
If I could get the funding I would create a BMS style commercial/GA sim, with a proper flight model. Eye candy is one thing, but what is the point when the aircraft flies nothing like reality?
That is it in a nutshell. As you can tell, I appreciate the fidelity of the BMS flight model.
One of my family is a test pilot, hence my obsession. I fly gliders IRL. That really gives you a feeling for handling.
Compare with BMS flight performance. I'm sure if you did these maneuvers at these altitudes in BMS, you would not make the airfield. Note when he drops the gear, too.
F-16 in FSX - Engine Out Approach and Landing:http://www.youtube.com/watch?v=2Ecex9B4zNc
I can't comment on X-Plane as I don't have it, but it was good enough to prototype the Virgin Global Flyer on. I know it uses blade element theory.
In summary, FSX is useful as a procedural trainer when coupled with realistic add-ons, but forget trying to do anything beyond straight and level with it.