I am not sure if the original poster is referring to the use of GUI file per test, or Global Gui file, or if they are referring to the ability to display these modes int he GUI Map Editor.
chilu - it just dpeends a lot on how you ultimately design your automation suite. In some instances, the ability to have a single GUI file that covers all objects of your application is the best, when your application isn't complicated enough to have overlapping object names.
Im my experience, the more complex the application, the more the need for a GUI File Per Test. It links in specific GUI Maps more closely to the individual tests. Also, as your GUI mapo grows in size, the slower the response of the test can be having to search for all the objects.
Then there is the final method Mercury tends not to really talk about, and that is designing your own framework to handle which GUI Files you need to load for each test/ windo you interact with.
As with either of the above, it all depends on your application, the experience of the automaters, and what works best for the planned testing objective/time allowed.
Insanity: doing the same thing over and over again and expecting different results