It references the following study
Old / Expired Link
$30 to view it :-(
The article comes from the perspective of gambling addiction, so I'm not sure whether it will give software names and so on. I believe statistics is Psychology 101, so presumably the quoted figure, that 'The results of the test showed that 39% of the sites paid out more than 100% during the demo sessions, clearly an inflated rate. ' has been properly verified.
Without having seen the study, though, it's very possible that 39% of sites could fairly pay out over 100%, depending on the game played. For instance, if you bet $100 (play credits) at 100 sites on 'odd' on roulette, then you would expect to pay out over 100% about 48.5% of the time.
So without seeing what the article says, a statisitc that 39% of sites paid out more than 100% seems pretty reasonable to me - casinos might have the edge, but without variance giving players a chance to win as well as lose, nobody would play.