Since the result cursor is quite small, I'ld create that first, with bKbKey added as column.
Then get a cursor of distinct bKbKey values
then calculate the aTotal field for that cursor and make a final select combining both cursors from there.

Might make a difference in cache behaviour....

select A.Field1, B.Field2, C.Field3, D.Field4, B.kBKey ;
  from      AFile A  ;
       join BFile B on B.kBKey1 = A.kAKey ;
  left join CFile C on B.kBkey2 = C.kCkey ;
  left join DFile D on B.kBKey3 = D.kDKey ;
 where A.AKey = ?gkKey       ;
 order by C.Field3, D.Field4 ;
 into cursor T_res

select kBKey from T_Res group by 1 into Cursor T_Key

Select E.kBKey, SUM(E.Field5) ;
   from EFile E ;
   Join T_Key on E.kEKey = T_Key.kBKey ;
   Group by 1 ;
   into cursor T_SUM

select Field1, Field2, Field3, Field4, aTotal ;
   from T_Res ;
   Join T_Sum ;
   on T_Res.kbKey == T_Sum.KbKey

perhaps create the T_RES already with the aTotal field and Readwrite for a later update.
At least you can get a feel which step costs time if you build distinct steps.

But those steps probably will always give you a perf hit.

my 0.02 EUR

thomas