Not entirely sure what you mean. Adaptivity is the process whereby Corona smartly decides which part of an image are more noisy than others and so could use more processing, and this process happens whether you use Noise, Time, or Passes as a limit (unless you turn it off :) ). [EDIT] This is what is visible in the Sampling Focus element, where you can see where Corona has decided to apply more processing power. [END EDIT]
If you set a Noise Limit of 5%, and it takes 25 passes to get there, then if you set a Pass Limit of 25, once that finishes you will have a Noise Level of 5%. Of course Noise is re-evaluated every 5 passes by default, so if you set a Noise Limit of 5%, and from pass 21 to 22 passes it would have gone from 5.01% to 4.98% (if it had been calculating the Noise in the image at that pass, which it wouldn't have been by default), it will keep right on going until pass 25 when it re-evaluates noise and finds it is 4.93% which is less than 5% and stops. But then 25 passes would still give you a Noise amount in the image of 4.93%. But if you really wanted to stop bang on 5% for some reason, you'd have to have only used 21 or 22 passes to get as close as you could to that 5% amount ;)
The Noise level itself is a blanket measure across the whole image (separate from Adaptivity calculating where to put most of the processing power), so if you have a really noisy cube in a tiny part of the image where the rest is just HDRI background, the Noise for the image as a whole will be very small because most of the image (the HDRI) is at 0% Noise.
So there are kind of three things in there about what you ask, and not sure which one you mean :)
