Originally Posted by
Lonewolf_50
Typically, airlines do not do actual stall training in the aircraft (valid risk and cost reasons), and do not have simulators that can accurately emulate a stall (lack of data points, apparently).
The simulators can emulate stall conditions to a fair degree of accuracy - certainly enough to train the points you refer to (nose-up, lack of roll stability, rapidly unwinding altimeter). The issue is that the aircraft behaviour in a stall is based on extrapolated data rather than actual acquired data. Judging by the simulated flight path in the report based on the PF's apparent inputs, the accuracy of the simulator is close enough that the difference is negligible, but it cannot be claimed to be 100% accurate.
I think the reason for focusing on stall avoidance was less to do with a lack of simulator fidelity than it was the industry-wide perception that by drilling stall avoidance procedures into crews, the need for stall recognition and recovery learned at PPL level need not be revised. This turned out to be a significant mistake with hindsight.