Just thinking here, but don't you pretty much have to create the graphics file that points the game to the correct graphics for the thing getting displayed by hand? IE, you have to tell the game that the graphic for the king resides in coordnate position 1:4 or something.
How about a utility that automatically generates said code?
Thinking of the top of my head, I'm envisioning an excel like program. Each grid location represents a grid location in the image. The image would be displayed side by side to the grid. The user then looks at the picture and can quickly tell the program that the image in 0:0 belongs to a miner. The image in 0:1 belongs to a king. Etc.
Once you finish filling in the blanks, the program would spit out the relevant code. For easy debugging of errors, you might even have a 'side by side' comparison mode in which the program extracts the 16x16 image, blows it up a little (like a thumbnail in reverse) and then displays the relevant information pointing to it. For example: It shows the image you made for the king, but the text file is saying the masonry job is pointing to it. Basically a proof reader, not really necessary, just fluff. (You wanted ideas, right?).
Maybe even make it simpler than that. Maybe have it import the image you are working on, blow it up a little, and then the user just fills in the grid, and then clicks the image that it points to. For example, I type in 'miner' in the 0:0 grid position, and then just click the miner graphic on the image, and it associates the two.