That book has been an inspiration for me and this is exactly what I am intending to do. However, it seems that the returns Clyatt used in the book (probably based on historical data) are much higher than what M* is projecting for the future.
Yes the returns are historical, but they are a necessary blend of very long term historical returns for some of the core asset classes and shorter term (20-50 year) returns for other asset classes like REITs or emerging markets, where the data simply hasn't been available.
The new edition will use the Dimson Marsh Staunton long term international data series, which actually lowers success rates a little, mostly because the European stock markets didn't always fare all that well during the 20th century, (to CFB's point).
Still the WLLM portfolios perform better than the M* averages you quoted mostly, in my view, because they bring in the additional asset classes and asset class permutations which boost expected return and contribute to an overall lower portfolio volatility -- Oil & Gas, REITs, Commodities, Private Equity. along with international bonds and stocks, as well as small stocks, domestic and international.
You can gain some comfort in the wisdom of this by the fact that big institutions today who make annual grants all seem to be investing this way, and making grants of around 4-5% of portfolio value.
By the way, the way the Work Less Live More SWR works is that you take, say, 4% of portfolio value each year, not 4% of Year1 value, with annual inflation adjustments. Real spending each year will therefore fluctuate up and down with market performance, but the result is greatly enhanced security (measured in survival of the real value of the portfolio) over long periods of time.