Don't really disagree. Where the cause-effect model usually breaks down in my experience is when you're trying to cater for abnormal conditions - which, given the degree to which simulators are used to train for failure cases, is quite important. Then the fact that light 'B' is lighting not in response to pressing the switch, but in response to some system operation dependent upon the pressing of the switch becomes vital. The elevators don't move to the position commanded by the column UNLESS the flight control system is working properly, etc.
The "simulated" versus "stimulated" debate for hardware is an interesting one. At first sight it would seem to be self-evident that it's better to have a stimulated piece of hardware, rather than a simulation of it's behaviour. But there are constraints - like transport delays, and assuring that the stimulation is actually valid - which sometimes make simulation a better approach.