This
http://www.innovationsforesight.com/...r-is-too-much/
gives some interesting rule of thumb estimates for guiding accuracy required for rough star roundness, i.e. 0.5" RMS for 2" seeing.
In general, a back of the envelope calculation shows that if you want to limit to an x% increase in FWHM in any direction, your RMS guiding error should be no more than about sqrt((1+x/100)^2-1)/2.4 times the seeing. So to limit to a 50% increase you'd need RMS guiding at about 0.5 times the seeing (1" in the above). 10% would require 0.2 times the seeing (0.4" in the above).
I'd say at 0.45" you're pretty good, Suavi!!
Personally, with my EQ6 (belt modded), I got the best I've ever got the other day, and that was still about 0.7" RMS. I'd be ecstatic to get a guiding graph like yours!!
Not too sure why I can't get better.