Well, the Shiller data set (what cFIREsim uses) actually uses GS10 bonds, not a "total market"... and I would think that giving cash a 3-3.5% growth number would be generous enough. I limit the data years to 1972-present, and still get pretty brutal results. *shrug* I'm just trying to understand the theory behind this, but I personally would never invest like this.
That link is blocked at my office, but I'm pretty sure I remember reading about PP using cash as a buffer during bad times. I'm just not convinced of the whole "cash buffer" strategy. There are a lot of reasons to discount the idea, mostly having to do with the drag that cash puts on a long-term portfolio.
Read the link when you get home. It's informative, and specifically addresses cash as well.
The Permanent Portfolio exclusively uses 30 year treasuries for bonds, and they perform quite a bit differently than the 10-year ones. And the average interest rate on cash doesn't account for
when it is higher and lower (for example, T-bills peaked at over 15% in the early 80s). The reason I mentioned using 50% bonds and zero cash in Cfiresim is that it best simulates the PP "barbell" -- average 30-year treasuries and 1-year treasuries together and you get the general performance of bonds with a 15-year duration. That's much closer to 10-year treasury performance.
To really understand the PP you'll need to read a bit about how it works and why and not simply look at a cfiresim run (because of the shortcomings I mentioned, as well as a few more like how the PP drawdown method and rebalancing works). I recommend browsing the blog at the link I provided, or reading the book the author wrote.
BTW, I personally use the PP and plan to keep doing so in retirement but I don't think it's the only good retirement portfolio. From all my research it's an excellent option, but do what helps you sleep at night and never invest in something you don't understand.