What seems strange to me is, if this visual approach at San Francisco is considered to be so difficult, why did the accident crew not use a GPS back-up?
It only takes a few minutes to create something that will give you some magenta lines to follow that look the same as an ILS, for a back-up if you are worried about not being able to eyeball a visual approach. It's just a matter of picking information off the approach plate to make up a pseudo-ILS using the GPS. Of course you would not use this as a primary approach aid, but where's the problem with using this as a back-up in visual conditions?
Even with the naked eye one would hope to be able to make a very close guess at losing 300 feet per nautical mile on final, which equals a 3º gradient. The rest of it would seem to be having the aircraft configured and on-speed prior to starting the final approach at a pre-determined altitude and distance, so that I don't understand where the difficulty lies.
On the other hand, not to notice that the speed is unwinding... that reads like a real lack of basic handling skill, as if there were a total dependence on automation to take care of one of the basics, speed control: the auto-throttle in this case that was, unfortunately, not engaged. Or in cultural terms, perhaps someone noticed that the speed was unwinding but felt himself unable to point that out.
I am not familiar with SFO. Is the visual to that runway steeper than 3º, and if so, why?