It's reduced by blinding because both groups think they're getting treatment - that is, they both get the placebo effect, and the test group also gets the benefit of the actual treatment.
Edit: It's not really reduced by blinding, it's removed as a variable because both groups have the placebo effect.
I suppose the fact that it's reduced by blinding the patient is a wash (consistent with both theses), but now you need to add two new causes to explain why single and double blinding work, whereas none of the observed phenomena are inconsistent with other well-known research biases.
Single blinding spreads the placebo effect to both groups since both groups think they're getting the treatment.
Double blinding reduces the placebo effect, since the researcher can't imply to the patient which treatment is actually real and which is placebo since they don't know.
Edit: It's not really reduced by blinding, it's removed as a variable because both groups have the placebo effect.