This is execution plan of my result query:
|--Compute Scalar(DEFINE[Expr1002]=[dbo].[MyGetAllOnLine2]([Mail].[CC])))
|--Hash Match(Aggregate, HASH[Mail].[CC]), RESIDUAL[Mail].[CC]=[Mail].[CC]))
|--Clustered Index Scan(OBJECT[MWolf_Mail].[dbo].[Mail].[PK_Mail]))
It is better plan that I saw. Unfortunately, specific UDF is pay for good performance.