When you add data tables from a data connection, you have the option to either keep the data table external or to import it into the Spotfire data engine and use it in-memory.

When you are importing cube data into Spotfire you will actually flatten the cube into a single data table, which means that some features that you have access to when working with an in-database cube are lost and you may see different results when aggregating your imported data in Spotfire compared to what you get when the external system does all the aggregation. Spotfire does not import aggregated data, only leaf nodes that then can be aggregated using the Spotfire aggregation methods.