SSOnly is for when you want to generate the extra Star-Schema table.
If you simply want to avoid aggregation, leave the aggregation clause of the Blend Unit dimension as Not Used. If that doesn't work for some reason, you can just set "YourTopMemberName;SSOnly", which will generate records only for "YourTopMemberName" and for the base members under it - as well as creating the extra Star-Schema table; you'd then filter out the top records at reporting time. (Note that if you set *any* option in Aggregation clauses, you have to start the clause with the top member of the slice of hierarchy you want to use; you can't just use qualifiers like "SSOnly" on their own.)
As for issues with your "one-member-dimension" hack: the problem with that is that it nukes any parallelization opportunity for BiBlend, slowing down things significantly. BiBlend parallelizes its workload by dividing it across members of the Blend Unit; no members, no parallelization. This is also why we tend to recommend to use the largest dimensions as Blend Unit, because they will generate more parallelization and hence complete faster.
