I found out the reason for the change in color in the calibration target in the NASA released color images is that in the images showing the Martian surface they used a near-infrared filter instead of the usual red filter.
This is from a conversation on slashdot.com:

=======================================
Re:If ya don't like their colors, then do it yours (Score:1)
by tommy_teardrop (228273) on Saturday January 10, @09:44AM (#7937278)
I've done a bit of experimenting with these files, comparing the colour wheel to the images taken on Earth, and in order to get a reasonable approximation of the 'true colours', you need the L4, L5, and L6 filters, for red, green and blue respectively. Ironically, the main problem is that there aren't yet pictures using the red filter. A lot of the images triplicated with the blue green and infrared images. (L6, 480; L5 530; L2 750)

The use of the L2 infrared filter means that the blue colours are oversampled (at least on the colour chart). Once the L4 filter images are used on the planet, full colour images will be ours.
-- IANAL, BIPOOTV
=======================================
Re:If ya don't like their colors, then do it yours (Score:1)
by rgclark (314480) on Tuesday January 13, @12:59PM (#7964175)
"I've done a bit of experimenting with these files, comparing the colour wheel to the images taken on Earth, and in order to get a reasonable approximation of the 'true colours', you need the L4, L5, and L6 filters, for red, green and blue respectively. Ironically, the main problem is that there aren't yet pictures using the red filter. A lot of the images triplicated with the blue green and infrared images. (L6, 480; L5 530; L2 750)

The use of the L2 infrared filter means that the blue colours are oversampled (at least on the colour chart). Once the L4 filter images are used on the planet, full colour images will be ours.
-- IANAL, BIPOOTV"

This still doesn't explain why the image of the sundial has the colors looking perfectly fine including the green and blue.
It's only the images that also show the surface or sky that wind up with the wrong colors for the calibration targets on the sundial.

Are you saying this image of the sundial was taken without using the red filter:

http://marsrovers.jpl.nasa.gov/gallery/press/spiri t/20040108a/PIA05018_br.jpg

Bob Clark
=======================================
Re:If ya don't like their colors, then do it yours (Score:1)
by tommy_teardrop (228273) on Wednesday January 14, @07:33AM (#7972065)
No - these calibration images are taken with the L4, L5, L6 filters. There are images of the color wheel in all the filters (L1-7, R1-7), but very limited ones of the surface/sky in anything but L2, L5 and L6. In fact, there are now L4 images available. The true colour sky looks white, but does seem to have a greenish tinge to it.

You can try this yourself if you have an imaging program that lets you use RGB colour channels. If you put color wheel images using L2 in red, L5 in green and L6 in blue, and move back to full colour, it looks nothing like the images you linked to. Blue looks bright pink-purple. If you use L4, L5, and L6, the colours come out just like that image you showed. Most of the posted colour images are colourised using the three main imaging filters (L2, L5, and L6), because those are the mainly used ones. It makes sense, the image quality is better, but one down side is that it doesn't give you as close a real colour as the L4 filter does.

I urge you to try it, as it's great fun... but then I'm a dork.

-- IANAL, BIPOOTV
=======================================

I can understand that with scientific observations the near-infrared might be very useful, but it seems to me for an image to be released to the news media for public relations the red filter would be used to give the most accurate color rendition of the surface.
I've also read that it's impossible to get a true color rendition even by altering the color ratios when this near-infrared filter is used.
In the last post copied above, the fellow suggests there are now some images that use the correct L4 filter for the RGB color combination.
Perhaps he is referring to images like the one visible here during this Spirit press conference:

You'll notice the sky is a very pale blue. Actually if you look at the mirrored sky in the calibration target image below you'll see that when the Sun is high in the sky the sky color is light grey to pale blue, similar to the way it looks in this image. This to me is itself a very interesting scientific fact, the sky looks light grey to pale blue when the Sun is high in the sky.
It is only when the Sun is low in the sky does the mirrored sky look pale orange in the calibration image. This is quite analogous to what we see on Earth during sunsets.
In short, I think it was a poor decision to use the near-infrared filter instead of the correct visual light filter in composing the color images to be released to the press to give an idea of the actual colors on the Martian surface.
(As I write this it occurs to me how strange that sounds, "You use an infrared filter instead of a visual light filter to give a VISUAL light image of the scene?")

Bob Clark