I saw a big change between VFP 3 and VFP 5, where the optimiser turned into a pessimiser for one of my queries. An internal change somewhere made my query over 20 times slower.
Rewriting to use a JOIN with FORCE to evaluate in the order I specified gave me the speed back.
Since the SQL engine has been updated, you may find SYS(3099,70) which puts back the VFP 7 behaviour could improve things.
On the other hand, somewhere I have a document from Joe Celko on how joins are handled in SQL92 and that might have a relation to what's happening here.
Also, I wouldn't necessarily assume that the beta is going to be fast just because VFP 7 beta was fast! The large changes in SQL may mean that there is more debugging code on there this time that wasn't there in the VFP 7 beta.