id like to avoid using an alpha chanel because it prevents one from useing the 8x compression advantage of the dds format (using dxt1). using other dxt formats decreases the compression ratio. if the glow map rendering code was altered in a manner that it would use a greyscale image rather than a color image, then you could combine glow/spec/env map into a single 24 bit image file. that way your texture would be in one map, and your effects would be in the other. usind dxt1 dds files, both the maps file and the texture file would still take up half as much memory as a single tga texture. the downside to this idea is that it would require a total overhaul of the glowmap code. in addition it would render any maps made by the community useless. of course all one would have to do is convert their env, glow, and shine maps to greyscale, then paste them into seprate channels in 24bit image in photoshop.
as for making the glowmaps greyscale without loosing light color data, it would be tricky. the way i understand them, they are simply rendered fullbright a pass after the texture gets rendered. instead would it be possible to multiply the pixels in the texture map with the greyscale glowmap, then render the resulting pixels as fullbright. this sound simple but it might be too much overhead when it comes to rendering effiecency.