For aggregation purposes should each query subject have a determinant that uniquely identifies the data aka the primary key? This determinate would have a check mark on both uniquely identifies and group by. Then the data should aggregate correctly for all columns that need to be aggregated? Of course this would only be when defined in the documentation under "When to Use Determinants".
In the dimension query subject, you would uniquely identify the column which is at the lowest level of granularity (based on the fact table). This would be the only uniquely identified column in that query subject. The other query items would be set to "group by" because they do not uniquely identify the rows in the fact.
in other words, all query items, other than lowest level granularity, will be group by while lowest level granularity would be uniquely identified