In general, checking identity constraints makes the task of document generation much more complex. (I believe I saw a claim once that it makes it NP complete, but I may have misunderstood.) So I doubt that you are going to have much luck telling Oxygen's example-document generator to obey the constraint you have in mind; most generators content themselves with producing data the satisfies the content models.
You may have better luck writing either a filter to distinguish examples that obey the constraint you have in mind from examples that don't (so you can generate 100 examples and run the filter to find the ones that are usable for your purposes), or a transformation to take an arbitrary example and change it to satisfy your constraint.