I think I know the answer, I tried the MDX in pivot4j-analytics application and got:
You need to put at least one level or measure on Columns and Rows for a valid query. You can drag elements from the cube structure and drop them on either axes in the pivot structure view.
So there's a restriction to have both columns and rows in the query. This shouldn't be an issue since 95% of the queries have both axes, but still it would be nice to see this included down the road as this functionality is supported by Jpivot.