Well it might be fixed in the next SQL Server version.
If not, depending on the record set size you could move it to the VFP layer and scan through it there.
Moving large amounts of data, even in ADO recordets instead of XML, isn't as fast as VFP :)