I think I unwittingly gave my response to this thread over in
another thread, but since it's probably more relevant here, I'll rewrite it here and add on:
I believe is the main (only?) reason to hold bonds in a portfolio for an early retiree is to give you something positive in your portfolio to mentally cling to in order to keep you from selling your stocks at a loss during a crash.
If you can keep yourself from bailing out of stocks during a crash (and if we assume the future is no worse than any year in the US past), a 100% stock portfolio is less risky over the long-term than a lower stock-allocation portfolio.
But, holding 100% stocks makes it less likely that you'll be able to stick to that allocation in a crash, thus likely making it more risky in real-world applications where human emotions are involved. That's why I like something like an 80/20% portfolio. It doesn't perform much worse than an idealized, robotically-controlled 100/0% portfolio, but it probably performs a lot better than a real-world, human-controlled 100/0% portfolio.
So, as a still-working 36 year old, I'm about 85/15%, and intend to keep the same allocation in retirement. At the moment my expenses are about 2.5% of my portfolio, but I'm still working because a) I'm a wuss, and b) a recent entry into a relationship has thrown much uncertainty into what my future expenses will look like. A conservative adviser might suggest that since I've already at least partially "won the game" to shift more of my money gained from the giant run-up in the stock market into bonds, but that would be for emotional reasons, since at least the simulators don't support that.
In cFIREsim you can leave the defaults (4% withdrawal rate) and do "Investigate Allocation" for various time periods (30 years vs. 50 years). The resulting graphs show that for 30 year periods you can reduce your stock allocation to 50% without much reducing your idealized success rate, while for a 50 year period you can only go down to 85% before the dropoff starts.
I'd guess your grandfather is biased by his age, and his personal unfamiliarity with the longer timeframes and significantly altered math that comes with early retirement.
Though it's possible he could have a more-sophisticated argument: "I know what the simulators show, but I also know the likelihood of you sticking with your 85% stock allocation over 50 years (as the simulation requires) is quite low, so thus your real-world success rate is actually greater with a higher bond-allocation than the simulation shows."