I see your point about the lack of controls in the "reparative therapy" here, but I think the reporter is making a very good point. With no controls, we have no basis for making the inference that the treatment caused change. The problem though, is with only retrospective reports from self-interested parties, we have a very, very limited basis for inferring that there was change. Unfortunately, the single-group-posttest-only is about the most common design with "applied" interventions, including job training. For what its worth. For my money, not being able to be sure that there is an effect to attribute to some cause trumps being able to pin the effect to a particular mechanism, in any particular case.