The best way to get AS5 to "settle" quickly is to run full load for 30 mins, turn off for 30 mins (leave the pump running if you can, otherwise turn off for an hour), and turn on, repeat the load, off cycle. By the third time you cycle your paste will have settled about as much as it is going to. You may pick up an extra 0.5-1.0C over a period of about 4-8 weeks, but then again you also may not.
3C would be getting towards the outer limits of the range of an expected temperature drop between a Storm/G4 and an MCW6002, depending on CPU type and load, and the quality of the mount of each block against the CPU. If you're running on a P4, which presents "compressed" temperature deltas to the user, then looking for a 3C difference may be a little too much to expect because P4's in general report less of a temperature delta than what is actually occurring due to the location of the P4's on-die temp sensor.
It also depends upon the reliability of one's thermal sensors (P4's not being terribly good at all), and closely monitoring ambient and water temperatures to ensure that the comparison being made is as even as possible at your end. Would probably need to remount a number of times to gain a good idea of the variance you are getting between mounts, and expected results between mounts, to be able to detect when you have a good mount, if you wanted to get the most out of the block.
A poor mount will affect both idle and load temps, in that both will be higher. You generally can't tell if you have a average or good mount by analysis of idle/load temps on a single mount. For terrible mounts you can, but when trying to pick the difference between good and average, you can't.