hi madhusudhan,
if you load data into infocube data is stored in F Fact Table. F fact table is default partitioned by Request id. The Request are different so it is not summerized and if you load in same flat file the data is summerized.
If you compress the cube, The data will be Transferred from F fact table to E fact table and all the request id's are deleted and data is summerized and you will get the expected summerized data.
Thanks,
Ramesh.