Key based image maps?

Has anyone been able to get key-based imagemaps working? The documentation seems to imply that just putting the frames of one's sprite in a row in an image, separated by the separated color is the right thing to do, but when I do this T2D doesn't acquire any frames at all. Deleting the leftmost column of separator pixels (so that the first column of pixels is the first column of my first frame) gives me a set of frames that are as tall as the whole image, and have appearently random widths (my frames are all different widths and heights).
Anyone have a clue what's going on here? Or an example of a working source image?
Thanks much.

-Leander