The (understandable, traditional) reason for measuring relative to placebo is that, "well, you don't need the drug if it's only as good as what a placebo gives -- just use the cheaper placebo! You should only bother with some special drug if it's better than what mere belief-in-effectiveness yields."
But that assumes that people are equally credulous of every placebo; if a person is more credulous of homeopathy, and credulity is hard to manufacture, then it's not quite wrong to say that it "works" under a relevant definition.
(I know this is kind of an angels-on-the-head-of-a-pen technicality, but eventually we'll want to actually understand why belief-in-effectiveness can matter at all, so we can exploit it, and reasoning like yours tends to give the idea that such an effort, by definition, can't work.)
To show something only has a placebo effect is the equivalent of showing it does not work.