What you’re proposing is effectively the same as "they should publish inaccurate guidelines that do not actually represent their informed views on the matter, misleading everybody, to pretend that they can prevent the stupid from being stupid." It defeats the very reason why guidelines exist - to guide you towards the optimal approach in a given situation.

And sometimes the optimal approach is not a bigger min length. Convenience and possible vectors of attack play a huge role; if

• due to some input specificity, typing out the password is cumbersome, and
• there’s no reasonable way to set up a password manager in that device, and
• your blocklist of compromised passwords is fairly solid, and
• you’re reasonably sure that offline attacks won’t work against you, then

min 8 chars is probably better. Even if that shitty manager, too dumb to understand that he shouldn’t contradict the “SHOULD [NOT]” points without a good reason to do so, screws it up. (He’s likely also violating the “SHALL [NOT]” points, since he used the printed copy of the guidelines as toilet paper.)