Consider the following applications, a finder application:
What is the user's task here? To navigate through the file-store in search of a file to operate on.
Now consider an open-file dialogue box:
What is the user's task here? To navigate through the file-store in search of a file to operate on: exactly the same. And yet the expression of these two tasks by the applications are completely different, with different presentations, and different interaction methods. A user must learn how to do the same thing twice, and can't use such hints as what the icon looked like, its relative position on the page, its date of creation and so on.