Characteristic relations will always be part of the Real-time cube, you can build your Aggregation Level on a multi-provider which contains at least one real time cube.
As by a planning function you will write/copy/delete... data from the cube the only the real-time cube supports these functions to be applied on the data.
Also note that when you have a filter it corresponds to the target cube on to which you want to write data.
"Char Rel/Data Slices/Key Date" can be defined on a Real Time cube and all the defined settings will be applied whenever you try to execute any function which writes/copy/delete/distribute... data on to the cube.