SQL> select num_rows,blocks from user_tables where table_name = 'ABC';

  NUM_ROWS     BLOCKS
---------- ----------
     20000         40

-- There are no indexes on ABC and so all access will be FTS
SQL> select index_name from user_indexes where table_name = 'ABC';

no rows selected

SQL> explain plan set statement_id = 'abc'
  2  for
  3  select * from abc;

Explained.

SQL> select * from table(dbms_xplan.display('PLAN_TABLE','abc','ALL'));

PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------

-------------------------------------------------------------------------
| Id  | Operation            |  Name       | Rows  | Bytes | Cost (%CPU)|
-------------------------------------------------------------------------
|   0 | SELECT STATEMENT     |             | 20000 |   410K|     9  (12)|
|   1 |  TABLE ACCESS FULL   | ABC         | 20000 |   410K|     9  (12)|
-------------------------------------------------------------------------

>> The rows (20000) identified by the optimizer now matches with num_rows <<

SQL> explain plan set statement_id = 'abc'
  2  for
  3  select * from abc where object_name = 'XXXXX';

SQL> select * from table(dbms_xplan.display('PLAN_TABLE','abc','ALL'));

PLAN_TABLE_OUTPUT
--------------------------------------------------------------------------------

-------------------------------------------------------------------------
| Id  | Operation            |  Name       | Rows  | Bytes | Cost (%CPU)|
-------------------------------------------------------------------------
|   0 | SELECT STATEMENT     |             |     5 |   105 |     9  (12)|
|*  1 |  TABLE ACCESS FULL   | ABC         |     5 |   105 |     9  (12)|
-------------------------------------------------------------------------

Now the optimizer thinks it needs only 5 rows even though it is a FTS because it arrives at that values based on this calculation:

1  select round(num_rows/num_distinct) from user_tables x,user_tab_columns y where x.table_name = 'ABC' and x.table_name = y.table_name
  2* and column_name = 'OBJECT_NAME'
SQL> /

ROUND(NUM_ROWS/NUM_DISTINCT)
----------------------------
                           5