Usability Testing: What gets tested, what gets usabilified?

Should usability research be performed through grounded theory--where results of a usability test determine how the product gets changed--or through traditional empirical theory (whether that is experimental, quasi-experimental, etc.), where the researcher begins with a specific question or hypothesis and tries to solve this through usability testing? What is tested? When testing in usability, is the item being tested? Yes. Is the user's ability to test the software being tested? Absolutely not.

How, then, does the usability researcher proceed to determine A) the product of the usability test (end result) and B) where the user fits into the usability test?

The product of a usability test should always be the tested item's improved capability to fulfill its intended function; how often should the intended function be adapted when users--who are very smart human beings, generally--use the item in an unintended-by-the-creator fashion?

Usability testing relies upon de-centering the user, giving the user no anxiety about how s/he performs, no leeway to consider whether or not s/he has a place in the product testing. Let me offer a sentence: "The user working with an item provides feedback on the effectiveness of that item's design toward fulfilling an intended purpose." "Working with an item" is the subject of this sentence--NOT the user; similarly, working with the item is the subject of a usability test, and combining multiple workings with the item provides results from which conclusions may be drawn (grounded) or that will respond to the questions asked.

I am fairly new to being on the testing side of usability testing; I've performed multiple usability tests as a user for various friends, and I've often discussed iterative design and usability in my work, simply because these are incredibly important issues in digital curation. But when I go to the testing side of things, I have to de-center myself, to recognize that I am not performing usability testing for any purpose beyond what the software is doing; the user and I (as the tester) are irrelevant and replaceable. What is important is the results of working with the item, and how I as a usability researcher structure those results back toward item redesign. There is not a third product that emerges from this researcher, for instance as a friend and I were discussing, improved software training; there is no room for a third product because usability testing is about the action performed with the item, and how combined actions contribute to an improved item.