I had previously done a few simple tests diluting a Photo Magenta to try and get a better color match in the grays, and it seemed that diluting it up to 50% didn't affect the color. However, these tests consisted of simply painting some ink on a scrap piece of photo paper with a Q-tip, after dabbing the Q-tip on a paper towel to remove most of the ink.
Like others, I have sometimes had a similar problem with getting proper feeding of the Cyan ink on a Canon i9900 (the first print was usually OK, but following prints looked like a progressively clogging Cyan head) and I decided to dilute the cyan to help this. I first diluted it by 10%, but when put into a small glass and swirled (like brandy) it still seemed to flow down the sides too slowly. I liked how it flowed at a 20% dilution (using distilled water), so that is what I used. When the refilled cart was first installed, I did a cleaning cycle and printed a nozzle check and everything looked good. I then printed (3) 8.5x11 sheets with a heavy color test pattern to further purge the old cyan from the print head, and found that I still had the same "clogging" problem. It was only after some playing around that I found that the vent on the cart was clogged (blowing into the vent did not push any ink out of the exit). This was cleaned with a quick blast of compressed air and the "clogged nozzle" problem went away.
Since the diluted cyan was already in the printer, I printed a color test pattern on 4x6 glossy photo paper to see what the colors looked like - all color management was turned off to get the "natural" performance of the printer/ink/paper combination. The test print showed that the "cyan containing" colors were now way off. A few more 8.5x11 sheets of the same heavy color test pattern were printed and the test repeated (to make sure that the print head was well purged) with exactly the same results. The same cart was then blown out and refilled with the original 100% cyan. The same purge process was repeated and the same test pattern printed on a 4x6 - the colors were now much better. Both test patterns were scanned side by side on the same scan, and the results are shown in "DilutedCyanTest". I also converted the image to CMYK and isolated the cyan channel, shown on "CyanChannel" in grayscale.
You can see that the diluted cyan did not affect the magenta, yellow, or red (which don't use cyan), but that it affected blue, green, cyan and black (all of which contain cyan). Not only did it affect the dark colors, it also affected the lighter colors in strange ways. For example, look at the second bar above the word "cyan". The diluted cyan even messed up the "white to black with cyan overlay" gradient, such that 30% point was lighter than 25% point! This can be seen better in the "CyanChannel" image, where most of the gradients containing cyan were now off. Look at the stronger color cast on the black gradients on the color image. These gradients are supposed to be pure grays, and the grays are usually the first place to show a color cast problem. As the color image shows, the diluted cyan changed a light pink cast in the middle colors to a strong red cast in the darker colors. The remaining color problems with the original cyan will be reasonable easy to correct. The color problems with the diluted cyan are non-linear and so bad that they can't be properly corrected even with a custom ICC profile.
To make a long story short - be very careful when changing the viscosity of inks with either glycerin or water, you may really screw up your colors. The ink supplier has to "get it right" - you will have trouble fixing viscosity problems yourself. This also points out how critical the manufacturing and Q/C are for the ink supplier - if your inks vary batch-to-batch, you will be forever fighting color problems.
http://www.nifty-stuff.com/img/files/DilutedCyanTest.jpg
http://www.nifty-stuff.com/img/files/CyanChannel.jpg
water even after shut off for several hours. Now, do I try the same diluted ink in the i9100? Hmmm.
