That indeed isn't correct...
We have never had a 'poorer' result.
But, there are a couple of initial points to be made based on a quick assessment of your uploads.
First, you say you used a 21^3 profile, but the data shows a 17^3?
(But that should still generate a better end result than colorNavigator.)
You also have not used Drift, which with a long profile is always preferable.
(That is exasperated by the relatively long integration time you have used.)
With Intelligent Integration you would be better using a lower value, so the overall profile is faster, with Intelligent Integration dealing with the darker colours - 0.75 sec. for example.
Also, with Intelligent Integration you would be better using Freq. or Period, as per the User Guides.
(AIO is not ideal, as it is a compromise way to use the i1D3.)
(Were the Colornavigator results using the same probe?)
Mob Boss at Light Illusion