My camera does not have an ND-filter, but as far as I know the bult-in ND has a limited range, and is used just to bring exposure within a "reasonable" range (i.e. a range useful to the standard usage a user would do of a point-and-shoot camera).
A 5 seconds daytime exposure is not within a "reasonable" range, so in a certain way you are trying to do too much for your particular camera?
A partial workaround could be to have multiple exposures and then to average them out. If you take 40 shots of a waterfall and then you average them out you will have an effect that is similar to a long exposure.
Similar, but not identical, though.... It should b ok for a waterfall, but surely not fo a long exposure of people moving: in this case you won't have a single, motion blurred image of a person, but rather several still images of faded-away persons, like some ghosts.
