Unlike in a diagrammatic model, there is no such thing as a connection occurrence - that is, if you create a matrix associating objects A and B and then create another matrix with A and B on the opposing axes, the connection between them is forcibly brought into view. In contrast to this, in a diagram a connection can be made between the objects and then if the same two objects are placed in another diagram, it is an option (but not forced) to bring the connection forward. Why is this important? Think about an As-Is vs To-Be model or one alternative vs another alternative. One representation might have the association and the other not (easy example - As-Is vs To-Be for an SV-1 - no interface now, interface later, but if the SV-3 matrix version is used to document this, then once the To-Be is created with the connection, if the As-Is is opened, the To-Be interface is brought forward). The gap is apparent in the APIs in that Connection Occurrence data can be drawn out of the repository for diagrammatic models, but it just doesn't exist for matrix models.
