Part of effective selection involves balancing off that which brings in a sufficient pool of talent to sift through, against the demands of how much it will cost, how much time it will take, and sometimes how much disruption it will create, to process and assess all those candidates. One of the devices a person can use to achieve that balance is certainly automated mechanically-scored selection tools that can whittle a mountain of applicants down to a smaller pile of candidates, but there can be a price to pay in relying too heavily on that.
As Terry notes, a more "authentic" RJP can be provided by simulations of the sort described in this thread, and referred to by Henry. While you can't depend too heavily on self-screening either, RJPs provide another means for whittling down the number of candidates to a more manageable quantity where one can now afford to invest the time in more labour-intensive assessment so as to make a better pick.
I look at selection as being like match-making. One is trying to find a good fit not only from the vantage point of the employer, but also from the vantage point of the individual. People have a right to be happy in their work, and if identifiable aspects of the job are not going to have you strolling in to work with a smile on your face every day, then it is only fair to let people know about them so they can focus their efforts elsewhere. I'm all for anything that avoids luring people into jobs that end up making them unhappy.