I don't understand the basic theory here. If we are dealing with two green channels that don't have the same color, you wouldn't expect them to be balanced, anymore than the red and blue would be balanced. Take a photo of a white object and the red, blue, G1 and G2 channels should have different averages.
That's right, and that is one of the reasons that most of us don't think that the two green channels are 'intentionally' different colours.
If the greens are the same, you wouldn't expect perfect balance...
I fully agree there.
There are at least one or two known reasons why they are more likely to be measured as at least a little 'unbalanced', than they are to be found as 'perfectly balanced'.
1. Small errors and tolerances in different amplifier gains.
2. Cross-talk/signal smearing in the read-out and A/D stages.
...Good maybe, but not perfect. Is 1% good enough? I have no idea.
I don't quite know at what level this is a problem either.
It appears that the problem mainly 'appears' due to particular 'demosaic algorithms' interpreting the differences between green pixels as image detail/edges, which essentialy exagerates the problem rather than smoothing it out.
I can make my old Minolta Dimage 7's G1 v G2 imbalanced several percent in either direction, just depending on whether the subject colour is red or blue - however I have never seen any maze artifacts in any of my Dimage 7 images - but this could be just because it doesn't use so called 'sophisticated' demosaicessentiallyexaggerates algorithms.
http://forums.dpreview.com/forums/read.asp?forum=1019&message=33359097
Surely Canon/Nikon/Sony/et al have a plan/methodology in place for any slight in-balance between any two channels, be they the same color or not.
Quite possibly, although it is probably done 'after' the RAW data stage.
One possibility is that there may be correction coefficients in the RAW file's meta-data ('Maker-Note' section) - and perhaps only Canon's DPP software currently makes use of it, but perhaps the beta version of ACR may not yet - but this is just speculation/theory.
You would think that any color filters would be less than perfect for amplitude and spectral response consistency. That would include the RBG1 and G2. I don't exactly see why the G1 and G2 would be "special" for any unbalance.
Again - I would agree with that. It seems very unlikely for the reason you describe, for the G1 v G2 filters to be different.
A misaligned CFA (Bayer filter) has been mentioned as a possibility, although this would be quite a stretch, as there is quite a clear gap between the photo-cells of the sensor, so the CFA would have to be quite badly misaligned to cause a problem.
