One of the challenges of twin training is accurately setting zero thrust. My feeling is most of the time more than zero thrust is being set and thus an unrealistic level of performance is achieved.
I did some training for a Navajo operator a while ago. The Chief pilot was sure that he could fly away from a low altitude engine failure and that the airplane climbed "just fine" on one engine. To disabuse him of his opinion we loaded the aircraft to gross weight climbed to 4000 feet AGL and actually shut down and feathered the engine. It was a big eye opener to everyone at just how little climb performance there actually was and as a result the SOP's were changed to require closing the throttles if the aircraft was not already clean and at blue line and even then be aware that the aircraft may still not have adequate performance and a glide landing straight ahead may be the best response to the engine failure.
The other problem with twin training is the engine failure exercise is always taught as a binary problem with only two conditions engine runs perfectly - engine has totally failed. In the real world a partial failure or surging engine is a more likely failure mode and is virtually never addressed in training.