I have taken a very quick look at one of the first Mark III images that Tom Droege has released. Since several people have asked about a "flatfield" vector for the image, I have created a quick version that may be of service to some.
I converted the "Orion2.dat" image into FITS format. Looking at the pixels at the extreme "left" edges (i.e. columns 1, 2, 3, etc.), I found that three consecutive columns had very negative values. The middle column had slightly lower values than its neighbors, with a mean value of -25699. I took this to be the "dark" value, and subtracted it from all the pixels in the image.
Next, I calculated the median value of the pixels in each column. This converts a 768-column by 800-row image into a 768-column vector. I contend that one can use this vector as a flatfield for drift-scan images. Here's a plot of the "flatfield" data values. The numbers on the left have a factor of 10^4 in them -- it might be hard to read the numbers.
As you can see, the center is more sensitive to light than the left edge, especially. The difference across the field is about 1500 counts, and since the peak value is about 13,000 counts, that means the raw images differ from "perfectly flat" by about 1500/13,000 = 12 percent.
If you want to get a copy of the "flatfield" values plotted in the above graph, just click here to download an ASCII file. You'll have to normalize the data values before you use them as a traditional "flatfield" vector.