Well that's the hypothesis - that since reversion to the mean stock allocation has always happened it will continue to happen. There's things that can invalidate the assumption, like the widespread common wisdom of HODLing index ETFs (till retirement, no matter what). If the hypothesis is false we could see stock market allocation go all the way up to 80%, or even beyond.
The author was not presenting it as a serious hypothesis. They were showing how compelling it looks at first glance, and how deceptive that naive analysis is, because of the unfair advantage the chart gets (in the sense that it inherently looks better than its predictive value). They did this to knock down many other valuation-based measures, not to advocate for this one.
Reversion to what mean, is the author’s point. The mean for the twenty years between 1900-1920 is different from the mean for 1900-1940, is different from the mean for 1900-1960, etc. The mean is continually shifting. All data sets revert to their own mean, on average, by definition. The mere fact that a data set does so is tautological and thus provides no evidence — no confirmation of any scientific hypothesis.
Edit: see my other comment for the quote of what I consider the key passage in the article.
Am I wrong?