You have to remember that all the electronics systems in modern fighter jets have been simulated extensively as has the air frame. Then someone gets to fly the jet to see if all that simulation was correct.
Is the simulation 100 percent correct every time? No. It depends on the models used. The models have to be good enough to use with the caveat that you could find a case where it's not correct.
Also bear in mind that even a $20,000 piece of test equipment could mislead someone, just like the simulator can. Case in point, if you swept a 60 db notch filter tuned to 77 kHz from 1 Hz to 1 MHz, don't expect to see the nadir. Hell, don't expect to see the notch. You'd have to change the sweep to a narrower band to see the nadir. I set my sweep samples to 9999 the maximum allowed in TINA. So using this example Tina would sample about every 100 Hz. I'd see a small dip in the signal, the same I would if I used an expensive spectrum analyzer. I first observed that sweeping from 50 kHz to 150 kHz while looking at a notch filter bank.
It's just another case of knowing your test equipment, and the simulator is just another tool in your toolbox.
There was s "paper" out on the internet that recommended a couple of changes to the 2n3904 transistor pspice model. There are a number of 2904 models out there, with some changes to each. I can find that link again if necessary.