I experience this same problem. I also use Paul-Marcel's method of calibration. When the rails sits flat on the table I can verify 90 degrees between the rail and fence. When I raise the rail via the guide supports and cut my stock, it is no longer square. I puzzled over this for awhile and finally came to the conclusion, as wopner (the OP) did, that the play in the guide rail support brackets (both front and back) were the cause. I can square everything up using multiple methods, but as soon as I unlock and move the rail up or down, the play allows them to come out of square when I lock it back down. Very annoying.
The Slop Stop won't fix this problem as the play is not between the rail and tab in the forward guide rail support -- it is in the guide rail support itself. (I use a tiny bit of duct tape over the tab and the rail has no slop in it. I can't imagine, however, if you add up the slop from the guide rail supports and the slop from the rail/guide connection how out of square the whole contraption could get! I know it's all relative, but the slop I experience, and the resulting inaccuracy of the cut, is outside the tolerances I need for my projects, which includes regularly crosscutting boards 12" wide or more.)
To overcome this problem I have resorted to checking the squareness every time I raise or lower the guide rail supports. First, I put my stock under the rail and against the fence. Then I lay the rail down on top of the stock. I then lay my framing square (which I have verified is 90 degrees) against the rail and then lean over to see if it lines up with the edge of the stock that is against the fence. When I am comfortable that everything is lined up, I lock down the guide rail supports. It always requires a little fiddling with the supports, pushing them to one side or other, to get to square. While not terribly difficult, it is annoying and slows down the process. It also makes the MFT less enjoyable to use, and as a hobbyist that's the reason I do woodworking!
I've tried to think of some solutions to this. Perhaps the tall Parf Dogs would make the alignment go more quickly as I could register against them when the rail is raised instead of having to fiddle with the framing square, but that would still require a step every time I raise or lower the rail. I could also fabricate my own guide rail support that would have no play, or at least less play, perhaps using some sort of sliding dovetail joint or even fabricating something out of aluminum extrusions. In the mean time, I'm just using the framing square to check calibration every time I move the rail.
The surprising thing to me is that this is the first thread I've seen that addresses this specific issue of slop in the guide rail supports, and I've looked on several forums and across the internet for some time. Certainly the OP and I are not the only people who have experienced this with the MFT? Or perhaps we just got a set of sloppy guide rail supports? At any rate, I am glad to finally see that I am not the only one who experiences this. Maybe others have some ideas to solve this problem that I have not thought of.
