Query using mathematical function of values from 2 tables has a performance problem

When I am executing a query that uses a mathematical function on values from 2 tables the query takes much longer than the same query that uses values from 1 table, even though the join remains the same.

Why is this happening?
Is there a way to bypass this problem?

Long query ( values from 2 tables ) :
SELECT
MAX ( ( SIGN ( attribute.keyValue- ( -2027587559 ) ) *SIGN ( attribute.keyValue- ( -2027587559 ) ) -1 ) *-1*data.val ) AS maxVal
FROM
DATA data,
ATTR attribute,
TREE_ELEMENT elm,
TREE_ELEMENT subject
WHERE
data.elmId=elm.id
AND attribute.keyValue IN ( 345647222,1569153803,1569146115,-2027587559 )
AND subject.id=elm.subjectId
AND subject.name = ‘test’


Short query ( values from 1 table ) :
SELECT
MAX ( ( SIGN ( data.keyValue- ( -2027587559 ) ) *SIGN ( data.keyValue- ( -2027587559 ) ) -1 ) *-1*data.val ) AS maxVal
FROM
DATA data,
ATTR attribute,
TREE_ELEMENT elm,
TREE_ELEMENT subject
WHERE
data.elmId=elm.id
AND attribute.keyValue IN ( 345647222,1569153803,1569146115,-2027587559 )
AND subject.id=elm.subjectId
AND subject.name = ‘test’


Long query execution plan:
Execution Tree
--------------
Stream Aggregate ( DEFINE: ( [Expr1004]=MAX ( ( sign ( [attribute].[keyValue]--2027587559 ) *sign ( [attribute].[keyValue]--2027587559 ) -1 ) * ( -1*[data].[val] ) ) ) )
|--Nested Loops ( Inner Join )
|--Hash Match ( Inner Join, HASH: ( [elm].[id] ) = ( [data].[elmId] ) , RESIDUAL: ( [data].[elmId]=[elm].[id] ) )
| |--Nested Loops ( Inner Join, OUTER REFERENCES: ( [subject].[id] ) )
| | |--Index Seek ( OBJECT: ( [TREE_ELEMENT].[TREE_ELEMENT_NAME_IDX] AS [subject] ) ,
SEEK: ( [subject].[name]=’test’ ) ORDERED FORWARD )
| | |--Index Seek ( OBJECT: ( [TREE_ELEMENT].[TREE_ELEMENT_APP_ID_IDX] AS [elm] ) ,
SEEK: ( [elm].[subjectId]=[subject].[id] ) ORDERED FORWARD )
| |--Clustered Index Scan ( OBJECT: ( [DATA].[PK__DATAS_SAMPL__485B9C89] AS [data] ) )
|--Table Spool
|--Index Seek ( OBJECT: ( [ATTR].[TREE_Z_IDX] AS [attribute] ) ,
SEEK: ( [attribute].[keyValue]=-2027587559 OR [attribute].[keyValue]=345647222 OR [attribute].[keyValue]=1569146115 OR [attribute].[keyValue]=1569153803 ) ORDERED FORWARD )



Short query execution plan:
Execution Tree
--------------
Stream Aggregate ( DEFINE: ( [Expr1004]=MAX ( [partialagg1005] ) ) )
|--Nested Loops ( Inner Join )
|--Stream Aggregate ( DEFINE: ( [partialagg1005]=MAX ( ( sign ( [data].[keyValue]--2027587559 ) *sign ( [data].[keyValue]--2027587559 ) -1 ) * ( -1*[data].[val] ) ) ) )
| |--Hash Match ( Inner Join, HASH: ( [elm].[id] ) = ( [data].[elmId] ) , RESIDUAL: ( [data].[elmId]=[elm].[id] ) )
| |--Nested Loops ( Inner Join, OUTER REFERENCES: ( [subject].[id] ) )
| | |--Index Seek ( OBJECT: ( [TREE_ELEMENT].[TREE_ELEMENT_NAME_IDX] AS [subject] ) ,
SEEK: ( [subject].[name]=’test’ ) ORDERED FORWARD )
| | |--Index Seek ( OBJECT: ( [TREE_ELEMENT].[TREE_ELEMENT_APP_ID_IDX] AS [elm] ) ,
SEEK: ( [elm].[subjectId]=[subject].[id] ) ORDERED FORWARD )
| |--Clustered Index Scan ( OBJECT: ( [DATA].[PK__DATAS_SAMPL__485B9C89] AS [data] ) )
|--Index Seek ( OBJECT: ( [ATTR].[TREE_Z_IDX] AS [attribute] ) ,
SEEK: ( [attribute].[keyValue]=-2027587559 OR [attribute].[keyValue]=345647222 OR [attribute].[keyValue]=1569146115 OR [attribute].[keyValue]=1569153803 ) ORDERED FORWARD )