Has anyone seen a better explanation?

Yes, that the original papers advocating the placebo effect were misleading in their reports and the popularisations thereof grossly exageratted.

Placeobo's can be shown to reliably have an effect on:

• Experience of symptoms.
• Even more so on reports of symptoms (that is, the presence of an expectant experimentor messes with people's heads big time.)
• Psychological state.
• Things that are significantly influenced by psychological state. The main two actual physical conditions that I can recall being genuinely altered by placebo (as opposed from being perceived to be altered) are ulcers and herpes virus (cold sores). Basically, two conditions that you more or less get from being stressed.

(I am not criticising the use of placebo controls here. But I am asserting that the primary benefit from such controls is in 'balancing out' other biases rather than because of direct effect of placebos on healing.)