I consider mental models more to be good decisions made in bulk (though not necessarily the best decisions). Of course you'll break the model when it's convenient or when you know better, but when you're taking your day on autopilot it'll keep you moving towards your goals. Like in the runner example, the mental model of "put one foot in front of the other" may not help, but a "I run for an extra 90 seconds after I get tired" might keep you pushing to always improve. Maybe this is just a difference of definition of "mental model". I saw it being complained about elsewhere, but the thread seems to generally be defining it as "rules of thumb".
You do have a good point that folks tend to have a lot of difficulty breaking mental models they used as crutches while learning. In playing guitar, I've found folks who rely heavily on things like the box model or pentatonic scale who plateau and get "stuck" at a certain skill level because of it. So it raises the question - Is it better to use a model that provides quick wins and a notable plateau, or risk never getting traction but having an easier time at the high level? Of course that all depends on the specifics of each scenario, but my rule of thumb is to do the more fun thing and re-evaluate when my skills plateau.
You do have a good point that folks tend to have a lot of difficulty breaking mental models they used as crutches while learning. In playing guitar, I've found folks who rely heavily on things like the box model or pentatonic scale who plateau and get "stuck" at a certain skill level because of it. So it raises the question - Is it better to use a model that provides quick wins and a notable plateau, or risk never getting traction but having an easier time at the high level? Of course that all depends on the specifics of each scenario, but my rule of thumb is to do the more fun thing and re-evaluate when my skills plateau.