It would be an obvious scam if it's true, it would mean the RTP level is different between the demo mode and the real one, or more concerning it could mean the game is not really random in one of this mode. I warmly invite you to expose this case here or in the Investigations section. On which game of which provider did your friends find a difference precisely? There may be a psychological bias when it is noticed by one single player, because he could have paid less attention to his losses in demo mode than in real one, but if several of your friends noticed that it's more serious.
I do not think that the demo mode would have a different RTP to be fair. However, when you make some tests, you are going to run only a limited amount, even if you run it a million times, it is not guaranteed. The code itself should be open source if you want to make sure but in that case anyone could search for loopholes to hack into stake as well so that would be a bad idea.
I believe that it is still the same, and doing your "testing" in the demo mode should not be a problem. At the end of the day, there is a house edge in both and if you think that you could somehow win in the long run, you will be wrong in any case and only kid yourself if you think it could be won.