Here's an illustration to show a bit clearer. For optimal result as much as possible of the red center frame should be used, and minimize the use of the overlapping segments of the green and blue frame. This makes the best use of the sensor's microlens offset.
Examining the overlapping segment, for example the one outlined in yellow, can be interesting. In a perfect system the colors would be exactly the same for both images, at least after white-balancing after a common point. I think the C1 LCC application can cause some white balance shift, so it may not match directly after LCC apply.
The sky seems to pick up some magenta cast towards the left side, I don't know if that's natural (depending on the sun's position the sky may not be 100% even) or if it's an artifact. Could be as simple as a white balance mismatch.
I know from previous testing that the sensor present very different color cast not only horizontal vs vertical but also if you turn the back upside down. This should be visible in your LCC shots. If the sensor behaves perfectly symmetrical the LCC shot from the green frame (left) should be the same as a mirrored blue frame (right), but I'm quite sure that's not the case, instead it will be more towards red on one side (the left?) and more towards blue on the other.
If the magenta cast of the left side sky indeed is a crosstalk artifact that effect may explain why it's not the same on the right side, on the left side you may get a bit more red/magenta, and on the right side a bit more blue/cyan.
There's a whole lot of speculation here though... I hope you'll share raw files at some point
The image does look great and natural. But say if the actual sky at the scene was indeed solid blue left to right and now it's not, is that an acceptable result? The greyer greens on the leftmost tree does not look unnatural, but if they in actuality was as green as the center bush, is that an acceptable result? For some it will, for others it will not. In the best case the greyer greens and the magenta left really was the look of the original scene, and that would be a lovely result. Unfortunately it's hard to nail down this for sure without lab-style testing, I have some results pointing towards that it should work, and others that are less positive. I don't really know what to think myself.
I
think that the greyer greens and magenta can either be natural or just a white balance error, and if so I think the HR40 will make most users satisfied. And after taking another look at Doug's library test on HR 32 in C1 and making a stitch "microlens-offset-aware" I'm a bit more optimistic.
Personally I would however not get the back without lab-style testing with the full range of wide angles I have in mind, I would just hate it if I'd been shooting a year and then realize that wait a minute, I have color shift in many of my shifted images that I just did not notice at first. I know MF users generally dislike lab tests, but for things like slight color shifts and desaturation which this is about it's much more reliable to make a lab test than shooting live scenes. A simple lab test would be to shoot a color checker in center, and then move the camera and shift around and shoot the exact same setup in various shift combinations, then overlay and look for differences. Then you would know what kind of color shifts and desaturations we're looking at for which colors, and then one can go on shooting live scenes and relate to what we saw in the lab. If I was a dealer I would do it first thing, the whole issue is so complex (the offset microlens thing doesn't make it easier) that I don't think it's really fair to say to a potential buyer "make your own real-world tests and make up your mind", because it's so easy to make mistakes of the kind that you don't really notice anything when you do your casual tests but then over time you notice that maybe there's something with the color not doing so well.
This thread shows how such a color-checker lab test can look (for a different back / lens combination):
http://www.getdpi.com/forum/medium-...osstalk-cancellation-preliminary-results.html
(unfortunately the early crosstalk cancellation algorithm demonstrated in that thread did not work out for the kind of crosstalk we see in this CMOS, but maybe I'll make another attempt...)