There are various techniques to eliminate the moment from the bed. An example would be to use a guide bar down the middle, kind of the opposite of the prusa that has the two guide rails on the sides and a central drive.
I realize that is something that is beyond the current hardware of the SM2… but it might end up being what’s required.
One way to help the Y axes travel out would be to loosen the screws tightening the modules down, move the Y to one end or the other, tighten the screws at that end, then move the Y axis to the other end and tighten the mounting screws at that end, then move to the middle and tighten the rest of the screws.That would help to remove any bias because the modules aren’t perfectly parallel. You might also want to loosen the screws holding the aluminum frame (keep them pretty snug) on just one side and then re-tighten them at the middle position.
Because the aluminum frame is not solid it can probably allow a small amount of difference between the Y axes. It sounds like a good exercise for my dial indicator!!
Just ran through the 5x5, this has been a game changer for me. Tried a print that just wouldn’t work for me on 3x3 even after extensive manual adjustment. After running the 5x5 it worked first time, not perfect. However, if this had been my first SM2 experience I would have been thrilled with that 1st layer.
Thanks again @Tone 
@Tone Yah really. Your efforts in this have been phenomenal. You have really gone above and beyond. You were always active during the SM, but at this point SM2 would be almost unusable for some people without your assistance.
@Tone I can for sure see using what youv’e started so first off thank you.
On to my situation. I’ve tried the default auto-leveling and then ran your check gcode and for sure there is a good amount of detla in the gaps across the 9 points. So now its on to your guide. Connected via USB to my A250 I manually sent the G1029P5 matrix command, then ran the “Level” macro that I copied from your “Level” above. The head did home, heat up then drop down to the bed and maybe start to scan, but during the process the nozzel embeded itself into the surface, luckliy almost at the edge where it might never cause me an issue.
Was that the very first point?
You current on firmware and Luban?
Because it’s so close to the top of the bed it makes me think that the sensor was off the bed and thus couldn’t sense the metal layer. Just my guess.
It should have started the scan close to x0,y0 and done a spiral inward.
@Tone, indeed that was the first point.
Firmware is 1.7.1.1
Luban is 3.4.2
Do a M420V right after doing the G1029P5 and post the results of the M420V.
Just the small mesh will do.
@Tone
Its worth noting between now and the time I made the divot I did do a factory reset of the A250, left auto-levling on, and ran through it. But here’s what I get now.
> G1029P5
G1029P5
others < Set grid size : 5
ok
> M420V
M420V
others < Bilinear Leveling Grid:
others < 0 1 2 3 4
others < 0 +3.488 +3.981 +4.291 +9.000 +9.000
others < 1 +3.483 +3.865 +4.199 +9.000 +9.000
others < 2 +3.346 +3.660 +4.011 +9.000 +9.000
others < 3 +9.000 +9.000 +9.000 +9.000 +9.000
others < 4 +9.000 +9.000 +9.000 +9.000 +9.000
others < Subdivided with CATMULL ROM Leveling Grid:
others < 0 1 2 3 4 5 6 7 8 9 10 11 12
others < 0 +3.48750 +3.65889 +3.83028 +3.98125 +4.09820 +4.19472 +4.29125 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000
others < 1 +3.49069 +3.64746 +3.80423 +3.94579 +4.06198 +4.16296 +4.26394 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000
others < 2 +3.49389 +3.63604 +3.77819 +3.91032 +4.02577 +4.13119 +4.23662 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000
others < 3 +3.48250 +3.61181 +3.74111 +3.86500 +3.97986 +4.08931 +4.19875 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000
others < 4 +3.44681 +3.56622 +3.68564 +3.80324 +3.91780 +4.03055 +4.14329 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000
others < 5 +3.39653 +3.50783 +3.61914 +3.73162 +3.84606 +3.96166 +4.07727 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000
others < 6 +3.34625 +3.44945 +3.55264 +3.66000 +3.77431 +3.89278 +4.01125 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000
others < 7 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000
others < 8 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000
others < 9 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000
others < 10 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000
others < 11 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000
others < 12 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000 +0.00000
others < leveling ON
echo:Bed Leveling ON
echo:Fade Height OFF
okThat looks right. The remnant 3 x 3 and it sets the other to 9’s.
What do you mean by “ran through it”?
If you try the macro again, change the feedrates to much slower,
Try changing the Y250 to Y240.
What it is supposed to do is go down to about Z10 then go over to InJn and start the auto routine.
Are you running the macro from the macros in Luban or from a file?
Went in and tried again, and this time it worked as planned. I should note last time ther was a loud beep by the system, i assume warning me something was wrong.
Need to go through the rest of the instructions now so I can do the minor extra tweaking, but the base seemed to go through through. New M420V
M420V
others < Bilinear Leveling Grid:
others < 0 1 2 3 4
others < 0 +3.875 +4.117 +4.432 +4.597 +4.702
others < 1 +4.020 +4.196 +4.449 +4.604 +4.755
others < 2 +4.000 +4.154 +4.372 +4.530 +4.679
others < 3 +3.917 +4.035 +4.269 +4.429 +4.574
others < 4 +3.809 +3.882 +4.134 +4.312 +4.416
others < Subdivided with CATMULL ROM Leveling Grid:
others < 0 1 2 3 4 5 6 7 8 9 10 11 12
others < 0 +3.87500 +3.95315 +4.03130 +4.11750 +4.22268 +4.33593 +4.43250 +4.50083 +4.55250 +4.59750 +4.63694 +4.66972 +4.70250
others < 1 +3.92944 +3.99962 +4.06980 +4.14824 +4.24532 +4.35066 +4.44134 +4.50629 +4.55658 +4.60255 +4.64617 +4.68547 +4.72477
others < 2 +3.98389 +4.04610 +4.10831 +4.17898 +4.26796 +4.36539 +4.45018 +4.51175 +4.56067 +4.60759 +4.65540 +4.70122 +4.74704
others < 3 +4.02000 +4.07593 +4.13185 +4.19625 +4.27838 +4.36898 +4.44875 +4.50778 +4.55597 +4.60375 +4.65444 +4.70472 +4.75500
others < 4 +4.02787 +4.08057 +4.13327 +4.19389 +4.27076 +4.35556 +4.43120 +4.48897 +4.53759 +4.58611 +4.63753 +4.68886 +4.74018
others < 5 +4.01741 +4.06856 +4.11972 +4.17806 +4.25093 +4.33100 +4.40338 +4.46074 +4.51042 +4.55958 +4.61043 +4.66075 +4.71106
others < 6 +4.00000 +4.04884 +4.09768 +4.15375 +4.22412 +4.30171 +4.37250 +4.42986 +4.48042 +4.53000 +4.58023 +4.62949 +4.67875
others < 7 +3.97810 +4.02290 +4.06770 +4.12106 +4.19101 +4.26952 +4.34111 +4.39863 +4.44922 +4.49884 +4.54915 +4.59849 +4.64782
others < 8 +3.94926 +3.98925 +4.02924 +4.07991 +4.15092 +4.23262 +4.30667 +4.36475 +4.41518 +4.46463 +4.51498 +4.56434 +4.61370
others < 9 +3.91750 +3.95236 +3.98722 +4.03500 +4.10704 +4.19199 +4.26875 +4.32810 +4.37926 +4.42875 +4.47819 +4.52597 +4.57375
others < 10 +3.88319 +3.91265 +3.94211 +3.98667 +4.05915 +4.14674 +4.22606 +4.28774 +4.34115 +4.39111 +4.43808 +4.48161 +4.52514
others < 11 +3.84597 +3.86970 +3.89344 +3.93458 +4.00747 +4.09777 +4.17991 +4.24461 +4.30115 +4.35181 +4.39536 +4.43303 +4.47069
others < 12 +3.80875 +3.82676 +3.84477 +3.88250 +3.95579 +4.04880 +4.13375 +4.20148 +4.26116 +4.31250 +4.35264 +4.38444 +4.41625
others < leveling ON
echo:Bed Leveling ON
echo:Fade Height OFF
okWhew. Glad it worked for you the second time.
Is it possible to just adjust the z offset later after doing a M500?
Adjust Z offset? No, not that I know of. In looking at the source the value is used just to offset the mesh. But,
you can use my Google Sheets to make an extra copy of the tweak macro and then adjust all the Q values to the increment you want. Don’t forget to do a M500 to save when you’re done.
This simple adjust would be a great boon to have as an option on-screen. To adjust z-offset in the eeprom, rather than just on the fly when starting a print every time. Having something stupid like a +/-0.01, and just be able to over time and a few prints really zero in. I wonder if that was the original idea behind the quick cal that used to exist. 
So I did the 1.8 FW update last night and ran the 5x5 calibration, so far I’m impressed, it does seem to be working better…although I’m thinking to get it dialed in a bit more having a program that would print a 20x20 box in each calibration location to after calibration would help us dial it in just a bit more. That way you could see if the calibration took or if a bit of adjustment via the command window by adding or subtracting to the saved values.
Dis you manage to use your glass bed with autocalibration ?
yes I did, very thin glass, but then I had an issue with the X axis stopped working and it then decided to break off the auto sensor…personally I hate them, I’d rather set it once and forget about it, not my first auto sensor to break the other was on another machine.