I think a lot of it depends on how you developed and verified this mental model. If you derived it and tested it with users, it will give the user a framework with which to understand and predict your system instead of requiring the user to wrap their head around your system, on your system's terms. This can be especially true if your system's information architecture is based off technical requirements or business rules.
A mental model is a cognitive construct of how a system works. It may or may not be accurate but it can be useful if it helps a users interact with the system successfully. A good mental model is one that seems natural to users and gives them enough of an understanding of the system to be able to predict how it will work. They are usually simplified or abstracted versions of the actual system model.
A more common term for a mental model is a "working understanding". Someone might not know exactly how their home heating and cooling system works, but he might have enough of a working understanding to be able to set it and maybe even troubleshoot minor issues.