Yes, I highly doubt that the Powerstrip method will ever compare to what we get with ATI drivers, where we're used to an almost deterministic behaviour. I'm not even sure if it's a good idea to fully implement this because it will encourage people getting the wrong video cards.
The fuzziness lies in the method used by Powerstrip for programming the videocard dotclock. We can only request PS for a given dotclock, and PS will produce the closest possible stable dotclock.
Part of the problem is that the dotclock granularity can show huge variations along the videocard's operational range, so you can have a lot of options for low resolutions but just a few of them for higher resolutions. We can't know this at first hand, so we have a problem if we wan't to create new modelines in real time.
This varies a lot depending on the video card, so while I can get excellent results with this laptop for any desired resolution and refresh (GeForce Go 7400), things are not perfect in other systems I've tested.
Second (and worse), is the fact that results are not always consistent for some cards and situations. So I've seen that some modelines work today but not tomorrow, and my guess is that it has to do with dotclock stability.
As for your case, you may get better results lowering yout desktop resolution to something closer to the ones that the vga monitor setting is picking. Probably around that resolutions PS finds more stable dotclocks to choose from.