SQL & PL/SQL :: Using Dbms_sql.execute To Build / Execute Dynamic Sql?

I'm working with old code that uses dbms_sql.execute to build/execute dynamic sql. In our case, the user can select varying columns(I think up to 20) with different where conditions as needed.

After building the sql, here's an example

WITH ph AS
(SELECT ph.* FROM po_header ph WHERE 1 = 2),
pf AS
(SELECT DISTINCT pf.order_id, pf.fund
FROM po_fau pf, ph
WHERE 1 = 1
AND ph.order_id = pf.order_id

[code]....

Where table records for

po_header = ~567746
po_fau = ~2153570

and PK "order_id" is a NUMBER(10) not null and a snippet of the code looks like

nDDL_Cursor := dbms_sql.open_cursor;
dbms_sql.parse(nDDL_Cursor, sSQLStr, 2);
FOR x IN 1 .. nCols LOOP
sCols(x) := '';
dbms_sql.define_column(nDDL_Cursor, x, sCols(x), 100);
END LOOP;
nError := dbms_sql.execute(nDDL_cursor);

why when the "execute" statement is fired off the elapsed time takes ~4.5 seconds but If I change "1 = 1" above to "1 = 2" it takes ~.2 seconds. If I run the above query interactively it takes ~.2 seconds. Shouldn't the above query when joining

ph.order_id = pf.order_id

return zero rows back instantly or does the "dbms_sql_execute" do some other type of parsing internally that takes cpu time.