Ah, I can see where the structure would work better for QlikView. Since it has it's own processing engine that builds a cube in memory (more or less). I did something similar when I used QlikView to get around the limitations of it not allowing multiple joins between objects in the structure (the whole circular reference thing).

Out of curiosity, why are you using Performance Point (PP) and QlikView? QlikView is more flexible than PP but I would think that maintaining both would be a pain. Are the QlikView loads filtered for a subset of your facts? I could see using it for departmental BI and PP for higher level stuff. We gave up on PP because Sharepoint was a pain to configure and we didn't feel that PP was flexible enough for our needs. We considered QlikView but didn't want to report from the relational layer since we put so much time into building out the OLAP. Unless anything has changed in the past 2 years, QlikView is not very OLAP friendly. So we went with Dundas Dashboard instead. It's worked out well so far.