My intuition makes me agree that it does address the rubber-hose scenario. Could you elaborate on why you think it doesn’t?
In some rather narrow cases it might help avoid random searches from a non-malicious threat. (e.g., border guard just proving a point for whatever reason)
If there is a purpose to recover something from your machine, an invested adversary that is willing to rubber hose is going to keep going. And the numbers seem to show that most people will not have the will and calm to use their duress password when they've been tortured into logging in.
You trick them into entering a password that either wipes the data, alerts someone to the attempt, or dumps them into a sandbox that appears normal but doesn't provide the access they want.
What do you think happens next?