OK, I'll post the corner's position that I've measured in mil. I've put the origin on the bottom-left of the board :
There is "corners" at (going counter-clockwise) :
(126,0), (3811, 0), (3811, 563), (3937, 563), (3937, 1331), (3874, 1341), (3874, 1537), (3740, 1537), (3740, 1635), (3937, 1635), (3937, X), (0, X), (0, 1562), (197, 1562), (197, 1464), (63, 1464), (63, 1228), (0, 1228), (0, 563), (126, 563)
X can be changed arbirtary for a variale size board. If X < 1635, then less corners can be made for a smaller board (such as NROM).
Holes coordinates :
(1972, 2665), diameter = 5.2mm = 205 mil
(2183, 3041), diameter = 3.1mm = 122 mil
There is also two "pins" in the plastic case that would prevent componant into going at some position on the board, but too bad I didn't measure their positions. I also haven't measured the plastic case's lenght for the maximum value of X.
Your version of the connector have slightly different version than mine.
I took a screenshot with yours (grey) over mine (yellow) so that the differences appears on the screen : (exernal link so it doesn't "enlarge" the thread)
Now, my dimentions are probably not "exact". I just measured them with a caliper or a ruler (for the longer ones) and converted the result into mils.
There is probably a range of values that will work, a range that will work pretty much but that will make you have to apply some force for the board to fit, and then some values when it won't work any longer. I have no idea if the error between both of ours versions is significant or not in this regard.
PS : It would be great if other people who sucessfully made PCBs (Memblers, bunnyboy ?) would compare their measurements to mine, so we could confirm their exactitude (however it might be possible that those people are hiding their info on purpose so that people buys their PCBs instead of making their own).