delete from my_table;

insert into my_table
select col1,col2,col3...
from ...
where ...;

commit;

1. you create a table
2. you load it with several million rows causing it to allocate 1 million blocks to the table.
3. all of these 1 million blocks have at least one row.
4. you scan the table (this will require looking at 1 million blocks)
5. you then DELETE all the rows in the table.
6. you then scan the table (this will require looking at 1 million blocks (even though there are no rows))
7. you then TRUNCATE the "empty" table.
8. you then scan the table (this will require looking at only about 4 blocks even though all 1 million blocks still below to the table))