KB484831: The calculation of metric with UseLookupForAttributes=false in a partitioned cube may return wrong data

This article describes the metric calculation behavior in the partition cube when the function parameter UseLookupForAttributes is false. 
As we know, if the function parameter UseLookupForAttributes =true for a metric, the aggregate from base/subtotal from the base metric calculates the attribute against the global lookup table. If UseLookupForAttributes =false for a metric instead, then the metric may be calculated by aggregating over duplicate rows.
Example: Suppose we have a metric [sumItem], defined as follows:   
Sum<UseLookupForAttributes=false>([Item Id]@[ITEM_ID])@{[Item Id],[Month Id],[Emp Id]} as [sum item]
In a non-partition cube, the original grid is shown below on the left. After removing the attribute [emp Id] from the grid, duplicate rows will appear on the grid. Because UseLookupForAttribute=false, metric will be aggregated across these duplicate rows and the metric values on the higher level are shown in the grid on the right.

In a partitioned cube, we expect to see the consistent behavior. However, prior to M2021 release with Engine version 13, the parameter UseLookupForAttributes is not considered when the metric calculation is conducted in the partition cubes.
 

1. Engine version 12 behavior:

                                                                                             
II.   In M2021 release under Engine version 13, we fixed this issue ( DE160932 ). The new behavior is now consistent with what we do in the non-partition cubes:                 
 

                                                                                                            
Note: This issue is fixed in M2021 under Engine version 13, and this change may bring performance degradation in some cases as we need to evaluate more rows in each cube partition.