Why would Anandtech base their conclusion on anything else besides the actual hardware they tested? It would be dishonest and misleading to do anything else. When retail cards with aftermarket coolers are released, you can bet Anandtech will review those, and their conclusion might be totally different than the reference conclusion.
The $100 price difference between the 290 and 780 is irrelevant to anyone who agrees with Anandtech that the noise produced by the 290 is unbearable for everyday use. It's like choosing between an average looking average income woman and a rich smoking hot chick with similar personalities. Every guy is going to pick the smoking hot chick until you reveal that she has this odd habit of stabbing you in the leg every night while you are sleeping. Suddenly all her advantages don't matter because no one wants to put up with getting stabbed every night.
I just felt Tomshardware of including the 290+cooler was a bit more thorough in its approach of making the two equal on a pricing point.
I do agree though that yes, the basic 290 with nothing done to it in reference form is pretty bad. But also, I feel like doing an open air test and using those results to say it's unbearable isn't something I'd have done. I mean, maybe they did use it normally, but I'd have at least liked to have heard "We used the 290 for a couple of days in a closed case system and the system noise was STILL TOO LOUD", or something of the sorts. A lot of people said the 290x was too loud based off reviews, but I've been talkign to a dude who bought crossfire the day it came uot and said the noise is perfectly fine, when you'd think it'd be unbearable. So at least some numbers in a closed system would have been nice just to even have not for comparison sake.