There is one other thing to consider regarding stall training, namely how applicants for pilot certificates (especially at the Commercial and ATP level) are evaluated on their checkrides, which feeds directly into how they are trained.
The FAA is currently in the process of revamping its certification process with the Airmen Certification Standards replacing the old Practical Test Standards for the "checkride" (i.e. oral examination and flight test) portion of the certification process. However, one of the major differences between the Private Pilot level and the higher levels of certification was that, at the Private Pilot level, the testing standards called for
"[recognizing] and [recovering] promptly after a fully developed
stall occurs"
Contrast with the evalution criteria for this maneuver at the Commercial and higher levels, where the applicant is expected to
"[recognize] and [recover] promptly at the “onset” (buffeting)
stall condition"
Now, if you've followed the second link, you will also see a note immediately below that states:
Evaluation criteria for a recovery from an approach to stall
should not mandate a predetermined value for altitude loss
and should not mandate maintaining altitude during
recovery. Proper evaluation criteria should consider the
multitude of external and internal variables which affect the
recovery altitude.
That addition is relatively recent. I don't have a PTS from circa 2007/2008 (when I was going through my own SEL and MEL Commercial training) handy, but I do not remember that section being there. In fact, the way evaluation criteria had evolved "in practice", was that any altitude loss during the maneuver was grounds for failure of that portion of the test.
On the one hand, that led applicants to be extremely proactive and initiate recovery the instant they felt a buffeting or they heard the stall warning horn going off. On the other, it also led applicants to recover by advancing the throttle to full and unloading the wing only the absolute minimum amount needed to avert an immediate stall, preferring a prolonged low-energy state over the altitude loss that a greater degree of unloading would have brought, even though unloading the wing more would have allowed the aircraft to accelerate out of the low-speed regime quicker.
Training is habit-forming (that's the whole point, after all) and that particular habit is now recognized to be dangerous enough that the FAA has both addressed it in the PTS and issued guidance to DPEs on the matter.
The Captain and FO on the Colgan flight, however, did their initial and recurrent training under the older way of doing things and likely picked up some habits that got them through their checkrides but where woefully lacking in a real-world scenario where they fell behind the airplane and found themselves in an actual stall, rather than at the "outset" of one.