If rendering to the same number of passes, it will not result in the same reported noise level for the final render, but will result in pretty much the same amount of cleaning up in the region (that region will look as good as it did in the render region). The noise level is calculated over the area being rendered, so you might get a different noise level from one region than another, and from a region to the whole image.
Best to visualize this with an extreme - consider a cube on a plane, which occupies just the center of the image (for argument's sake, let's say the cube and plane are in the center rectangle if you divided the image into 3 rows and 3 columns). The rest of the image is a plain white background environment with no geometry. Select a region of the image that is just background, no geometry, render for 5 passes, it will report zero noise. Select a region that is all cube and plane, render for 5 passes, and it will have some noise percentage. Render the whole image for 5 passes, and it will have some noise percentage, but a lower noise percentage than when you rendered just the region with cube and plane (since there is now a lot of the image that is background and noise free). But the cube and plane will have cleaned up by pretty much the same amount as it did when rendered as a region, since you rendered for a set number of passes.
The same applies in less artificial situations - that region of a room that happens to focus on the bed, with the shadows under it, and the detailed rumpled bedclothes will have more noise in it than a region with the window with just a simple frame and mostly background environment, or a region with a wall and door, so will report different noise levels for the same number of passes.
So in your case, if you have chosen the region well (that is, an area that is likely to be one of the noisiest in the image), and found how many passes it takes to get that region looking good, rendering the whole image for that number of passes will a) take longer, of course; b) clean up that area of the image pretty much the same; c) give a different reported noise value for the final render (probably a lower noise amount if you chose an area that was likely to be noisy, as the rest of the image will have less noise in it and so will lower the percentage of noise reported - basically, the easier to clean up areas will have much less noise in them at that number of passes, they've probably been rendered by more passes than they needed to look clean :) ).
