I would definitely choose a red filter over blue for NDVI. The comparisons I did in the fall suggest that using a red visible band for visible light in NDVI calculations does a better job discriminating between plants and non-plants. Ned's work convinced him of the same thing. It also reduces the haze problem and makes comparisons with legacy NDVI more interpretable. Using a blue band for visible in NDVI calculations (as in infrablue) can also be revealing, for example it seems to allow identification of unusual pigments that absorb at the red end of the spectrum, but mostly when you can compare infrablue with superred. The Wratten 25A gel filter I put inside the A2200 did not seem to cause much more optical degradation that the glass BG3 filter.

That histogram is from PS CS5, and is slightly different from CS3. You can find it under "Window."

Ned's plugin allows you to select the camera channel (RGB) to use for both the visible and NIR band in NDVI calculations. This is true in both the single camera and dual camera tools. So when using the standard two camera setup, you have three choices of NIR bands (from an NIR converted camera) and three choices of visible bands (from the unmodified camera). That means you can simulate the NDVI computed from single cameras and start to learn about the impact due to, e.g., the visible band from an infrablue or superred camera always being contaminated with some NIR (unlike in unmodified cameras). If you take simultaneous photos with both an unmodified and infrablue camera, you have two choices of visible blue band (one from either camera), and you can learn about the impact of the NIR contamination of the infrablue blue band.

Infragram.org does not have an option to process superred (e.g. Wratten 25) photos. The infragram sandbox originally had the capability to enter any formula for NDVI, but I couldn't get it to work in the new version. If you were really committed to using Infragram.org to process superred photos, you could use Photoshop to swap the red and blue channels (Image/Adjustments/Channel mixer). That turns superred into infrablue. To do this automatically, record an "Action" (Alt-F9). You could then use Bridge to process an entire directory.

It might be pretty easy to add superred processing to Infragram.org, but a higher priority might be to add histogram stretching. It's really hard to simulate legacy NDVI without it.