So while I don't dispute your link (because it's correct), the placebo effect is normally defined as the effect over and above that seen in the no-treatment group (who don't get a placebo or treatment). Regression to the mean would impact both placebo and no treatment (as well as treatment) groups equally, so cannot be the explanation for the entirity of the placebo effect.

Placebo patients know that they have received a treatment, so that introduces a bias in the reporting of outcomes from the patient, even if the medics are blinded. If you believe you have been treated it is well known that patients report improvements even for illnesses that can be measured not to have improved.

The breaking of blinding and contrary measurements is enough to undermine most claims for placebo.

Yes, I agree. That's what the no-treatment group controls for. And generally, both medics and patients are blinded.

I sometimes wonder about the validity of comparison with the placebo. What if the drug being studied makes the patients feel something, but doesn’t directly help the condition being studied. But that something feeling alerts the patients that they’re getting the real drug and not the placebo, and that awareness causes all the psychological biase we’re worried about?