If that's the point it seems like a particularly inaccurate and cherry-picky one. Security is not what's introducing the failure modes, the adversarial universe is. The no-security solutions have catastrophically worse usability in that reality. You're comparing the usability of things with security features with the usability of things without security in a hypothetical non-adversarial environment. I can posit a world in which security features never fail or introduce usability friction, that wouldn't be a useful starting point for a comparison either.