Alternative medicine and Statistics

A good article in this week's Economist Magazine summarises the overall results from one Professor Edzard Ernst's career spent testing the efficacy of alternative medical treatments. Overall his team has found statistically significant efficacy for 5% of all treatments studied. This was reported by the Economist as evidence that 5% of alternative medical treatments were "truly" effective.

But is this the correct interpretation of this outcome? Putting on our statistical hats, assume that NO alternative treatments were any better than placebos, and that Prof. Ernst and his team had examined their efficacy using a 95% cut-off for level of statistical significance. In what percentage of the studies would you expect them to (incorrectly) find that the treatment was statistically significant? (i.e. make a type-1 error) Ha ha!! Is that a coincidence??!