Behavior very strange

Question

I make a query of the Customer Table with Left Join to other two tables and when I put c.DocumentNumber like '1%' ó c.DocumentNumber like '112%' the query returns rows in one second and when I put c.DocumentNumber like '11%' the query returns rows in 3 minutes!!!!!!
What’s happens!!!!
I have an index with LastName and other with DocumentNumber
The query is more or less this:
SELECT c.Customer_ID,c.LastName
FROM Clientes c
     LEFT JOIN .......
     LEFT JOIN .......
WHERE rownum < 501 and c.LastName like 'GARC%' and c.DocumentNumber like '11%'
If I see the Explain Plan in TOAD:
When I execute c.DocumentNumber like '1%' ó c.DocumentNumber like '112%') the Cost is 1098, the time to return rows is 1 second and use the LastName Index.
And when I execute c.DocumentNumber like '11%' the Cost is 947, the time to return rows is 3 minutes and use the DocumentNumber Index...
What can I make to the query run faster with c.DocumentNumber like '11%'???
It’s normal?
What’s happens???

Answer 1

For the charge of the records in another schema I put "trims" in the fields because its have "spaces"...
Your questions:
* Did you gather the stats for this new table/indexes on another schema?
I analyzed the table and the indexes on another schema...
* Is it using RULE or the CBO?
I think CBO because have a value in the COST
* Can you post the explain plan for the query against this new table?
I post the explain plan for the NEW Table:
Operation     Object Name     Rows     Bytes     Cost     Object Node     In/Out     PStart     PStop
SELECT STATEMENT Optimizer Mode=CHOOSE          1           4                     
COUNT STOPKEY                                        
TABLE ACCESS BY INDEX ROWID     MASTERS.DIM_CUSTOMERS     1      330      4                     
INDEX RANGE SCAN     MASTERS.IDX_LASTNAME_1     2           2                     
ACTUAL SITUATION
* Now in the original schema I have only 10.000 records and If I execute now the queries "in the original schema"
Its are executed with the "correct indexes"...
Thanks!