That's not how that works. The chance of each toolhead breaking is independent. For this example two toolheads is only 79% confidence of getting at least one with six quenches. For a 95% chance of getting a six quench with 55% chance of success per toolhead you would need four rather than two, and there's still a one in twenty four of not getting it.
The follow up example is worse, you are not owed one working of ten attempts. In fact with 11% obtainability ten attempts has a 31% chance of not producing a single toolhead to the desired specification. In fact you would need twenty six toolheads for 11% obtainability to have a 95% or greater chance of producing a single functional toolhead. However, that's still one out of twenty one attempts failing completely.