10% isn't universal. It is a common number to use to reduce maximum loads to get to a safe start pressure. If you look in enough databooks, you'll find starting loads in one that are substantially lower than 10% below the maximum load in another. But you also find some data with starting loads less than 10% below maximum.
There is a problem that occurs when powders can't develop enough start pressure, which is the powder can extinguish, leaving a bullet stuck in the barrel. If the next round fires into that stuck bullet and doesn't extinguish, that can cause catastrophic damage. The exact load level that causes this varies with powder and bullet weight (heavier bullets improve confinement for building adequate pressure).
Also, some powders have normal loads at low loading densities in some cartridges, and if you reduce them too much, pressures can get erratic from the low loading density. Dr. Lloyd Brownell documented pressure having random scatter that included a wide range of highs and lows below about 60% loading density in 30-06. And this was with very common single-base stick powders like IMR 3031. Some of that erratic behavior is credited to the fact you can get a lot of powder surface area exposed to the primer flash when the density gets low, giving it a bigger head start on burning up. A lot of load data, knowing the density of powders change from lot to lot, recommends never going below 70% loading density because of this. However, it does not cause a problem with fast powders used in reduced loads, because they light up fast whether the flash covers a large starting area or not.
