Further, I found this article: Schema Modeler and Stored Procedures (again)
This seems to suggest that I can’t simply treat the schema table as a normal one, setting parameters and then opening it, but instead have to use different means to access the results.
If so then this seems nuts to me. Surely the whole point of DA is to abstract the database and the client shouldn’t care whether the data is being provided by automatically generated SQL, hardcoded SQL or a stored procedure. The schema has all the necessary information to execute the stored procedure with the correct parameters and return a dataset.