There's no need to invoke a placebo effect as an explanation for any of the results. He took it in response to acute stress, which is a local maximum in stress. The decrease in stress afterwards is to be expected.
Take a look at the stress vs day graph, though: it sure looks like stress was lower during that period. But maybe this is because the measurements were at a different time of day?
