The data structure rendered by the engine is a scene graph, that is, a tree of spatials. There are basically 2 kinds of spatials: (1) nodes (which are invisible but may reference child spatials) and (2) geometries (which are visible but never have children).
Oto is a model, a tree of spatials you can load from an asset and attach to the scene graph. The tree returned by loadModel() contains both nodes and geometries, but you reference it by means of the top node returned by loadModel(). If you load the same model twice, you effectively get two copies of the model, though I believe there’s some sharing of data.
Models are usually created using a tool like Blender, then converted to a more convenient format: the native J3O format. (Oto comes in Ogre XML format, so it’s not a good example of this.)