Ya'll covered it all (MadDog gets an upgrade to "Elite"
), except...
AS takes 48 hours to "settle".
Replacing the pump will help: although the head is actually pretty good, the flow rate is rather low, and the actual flow rate will be significantly lower, which reduces the performance of the waterblock (which WB is it, BTW?).
However, all things being equal, the performance should improve, with a better mount. In this particular case, because you state that you used less AS, I would expect an improvement either way (it's a big assumption, but I'll go with it, for the discussion).
There is a remote possibility that when the block wasn't sitting properly, it may have been sitting close to the CPU diode, throwing measurements off a bit, but that would be a real freak occurence. It certainly doesn't apply with an IHS.
It should be noted that lapping a P4 is a bad idea: although it can be observed that it is not flat, there's a reason for that: under the proper clamping pressure, it flattens out perfectly. That makes it extremely difficult to lap this CPU, because one would have to try to match the same curvature, as the nickel plating is removed, exposing the copper of the IHS (Integral Heat Spreader). This is, IMO, the most likely cause of your temp difference.
So what do you do? If you still have nickel on the corners of the IHS, you could try to continue lapping it, and make the nickel dissapear. If you don't have any nickel left, which would have guided you in the lapping, then you could try it, until you get it right, or just pop it off (at your own risk).