Non additive and additive measures

Hi,
I have a question, I designed a fact table which contains both additive and non additive measures in base table. Can we design a fact like that. I found that when we group the data from this fact the data is correct. However there is a requirement to summarize the data in a summary table, From this summary table data is again grouped. I found that the queries with different groupings on base table gives correct results.However grouping on already summarised data gives anomalies. The ETL team has told that the issue is with the design. I want to confirm that there is no issue with the design.

It sounds like dimensions in the Fact summary table are defined or loaded by ETL incorrectly. An easy way to find out to ask the question what is the grain of that summary fact table... in other words how does a single row represent in the Fact table. A Fact table with different grain would NOT produce correct result.
Try comparing base table's data to summary data to find out where are the differences.

It would help if you explained the design.

Generally you want to avoid semi/non-additive measures. What are these measures? Can they be expressed in another way to make them more useable?