> Is it really necessary to define the column AA as an Index to get it running? This would mean that all colums should be defines as Index.
> Or did I ignore some facts?
You need to check the plan and the documentation around "dependent joins". In situations where there isn't enough information to determine the relative size of the rows being fetched from FILE_0063 and FILE_0243, then we'll default to a standard join so that the source access can be processed in parallel. An index, cardinality/ndv stats, or a hint will choose a dependent join.