Well, it depends on the situation, really. I know that it's talking about what you said, and you're right, but, just bare with me.
If you're testing something that does a series of its own trials, and 60% of the time it does its set of trials correctly, then 60% of the time, it works 100% of the time.