Hi,
maybe this is a late reply, but also note that 4.725 alone already cannot be
represented in floating point exactly (and this has nothing to do with Postgres).
Just sum it up 100 times to "see" the round off error becoming visible:
chris=# select sum(4.725::double precision) from generate_series(1,100);
sum
------------------
472.500000000001
(1 row)
vs.
chris=# select sum(4.725::numeric) from generate_series(1,100);
sum
---------
472.500
(1 row)
Bye,
Chris.