As a general rule, reality always trumps a simulator.
If your circuit works in reality, but not in SPICE, guess who is wrong? Certainly not reality.
If your circuit doesn't work in reality, but does work in SPICE, guess who is wrong? Again, certainly not reality.
I agree with the latter, but not necessarily the former. A circuit that works in reality but not in SPICE may be OK, but that is a very good hint that you may have done something horribly wrong. A circuit that appears to work but relies on some parasitic element may stop working tomorrow, or with a new batch of parts. Technically of course the simulator is still 'wrong' in that its results are not the same as your measurement, but in reality the problem is with neither the simulation or the measurement but your circuit.
At the very least if you come up with a circuit that appears to work but simulates poorly you need to bang the hell out of it. Make sure it works over the full range of temperature, input signals, and loads, and with a selection of active devices to make sure you didn't just get lucky with the perfect device.
Really, if you bothered to make a simulation and it disagrees with reality you should figure out what aspect of your circuit isn't being simulated accurately. Then you should either fix the simulation, or have a coherent explanation of why simulation is not practical. If you aren't prepared to do this, skip simulation altogether, it is just a waste of time.