The detection of 16bit pixels was not working on my machine with Linux and open source Mesa/Gallium drivers for ATI, although GL_ARB_half_float_pixel was on the list of supported extensions. I checked the test and it appears that the prefix "GL_" is missing. After adding this and recompiling DftD all works fine and I don't get the warning anymore.
Proposed patch
Ping. Any problem with this patch?
Better late than never i suppose