You are forgetting about uneven bed where you will have like 0.5mm or even larger gap in the middle. In any case, if someone wants to use silicone for a whole bed, usual RTV silicones will not work as silicone in the middle won’t cure as it will be sliced between aluminum and glass which do not pass moisture required for curing. Magnet rubber will pass some but not that far. So only two component silicones will do the job. BTW there are some silicones with high thermal conductivity like 2 W/m·K but they are very expensive.
“The solution he ended up using of an extremely thin layer of 3m adhesive is going to be better than the silicone, and was a good pick.”
But it won’t fill any gaps, therefore not better than air unless both surfaces are flat.
Until I got some time to write a full article here are some first results with the mirror+magnetic sticker bed :
I have used this sticker & PEI plate for my tests :
Temperature :
About temperature, it is almost ok, about 5°c less than the setted value (measured on the first layer of plastic once printed, so the bed may be a little upper).
Flatness :
The bed is almost perfectly flat, but as it can only be leveled on three points, the back left and right are not at the same height, so the bed is a little tilted about 250µm
On all the area of the bed, the ABL widget only move from one graduation (250µm), and it is mostly du to the tilted bed
Magnetic sticker:
Very strong, at least as much as the Bambulab original, maybe a bit more
First layer with Giantarm PEI textured plate:
I ony do a 0.2mm first layer test on full bed and it come out nearly perfect, only two little defect that are probably due to the bed was not perfectly clean
I tried a 0.08mm first layer test on a small area (10cm) and it came out perfectly, I will make a try on full bed when I got some time
Bed height:
The bed is about 4mm higher than the original, it is a bit anoying because the little hooks at the back that help placing the PEI plate on the magnetic sticker cannot be use, I will print and stick new ones in PACF
But this is interresting has it enable to use PEI plate that are a little larger 260mmx260mm wich was not possible on orignal bed
Overall, it is IMO a full success
Bed temperature set to 60°c :
Measured the new bed with some good and suprising results.
The new bed was less banana (0,44mm instead of 0,86mm delta cold). A good start!
Then I followed the recommendation of heat soaking the bed for a day. So I printed some PLA with 35° and 55° bed temps. Inbetween I let the bed sit with 90° and 110° for 30 minutes.
Well, the banana became less of a banana. See for yourself. I consider this good enough in terms of what I can expect from a machine at this price point. What do you think?
… and when I find the time I will follow in DzzD’s footsteps. Until then some aluminium tape will bring it to 0,1mm delta
Encouraging news! Also, good work Frank. I might try the magnet upon magnet craziness this weekend 
0.1mm first layer test, wich turn out perfectly :
EDIT: It is pretty amazing to print out a such nice plastic layer that is as thin as a standard paper sheet, which is as much flexible and solid than paper
The only way that a single “Z” rod would be moving at a different rate is if the pulley was a larger or smaller diameter or tooth count. Highly unlikely. One motor, driven units with the same size pulley, 18 idlers and or guides or none and I guarantee you the driven pulleys are running the same rpm.
Now in the case of post DZZD made about one corner not being able to adjust enough, the rod is off by one tooth. Loosen the tension, rotate the rod in the needed direction 1 tooth and try again.
I am not sure, you can simply apply z screw manual adjustment : Manual Bed Leveling / Manual Bed Tramming | Bambu Lab Wiki
But the back is only supported by a center lead screw, that’s explain why left may be upper or lower than right
You would have to lay the printer on it’s back or side and loosen the “Z” tension as much as possible, then get all the belt slack to the one “Z” pulley that is off, and rotate it 1 tooth on the belt. IIRC it is a 12 tooth pulley so you would only be rotating it 30*, and that can be translated to the thread pitch (mm per thread or threads per CM) to tell you how much you are moving it up or down.
If it’s the rear corner that is off, then it could be that the “Z” nut in the bed may be at a little bit of an angle.
A straightedge across the back frame and/or the offending side will tell you if the frame is tweaked at the back causing a corner altitude difference, or if the “Z” nut is not perpendicular to the bed.
My first layer also looks good. However, I have slight gaps between the webs in exactly the same place.
Your photo also shows a different surface at this point.
Is this due to faulty auto-bed leveling?
3 screw adjustment is not sufficient for correct leveling of the far corners only if bed is not exactly flat. I suppose most of the warping is in diagonal direction. Otherwise there would be no tilting.
Correct, it would only change the corner that has a “Z” nut in it.
I dont think so, the picture is misleading, the layer not as much bad, it is ok, but there is a sligtly different appearence on top photo, dont really know why, maybe it is because I have touched the bed at this location before printing. The defect is not visible from other points of view.
EDIT: also the bed get cooler at this location, due to the tab
That’s strange. I have the problem with both the textured plate and the smooth one.
Could you maybe provide this first-layer file? I would like to print it on my printer.
Hi
Nice one!
As written many times before. The bed warping at heating is very small, yes it warps but almost negligible.
If you look for my posts, you will find my pictures of the outer frame. I do belive the reason you dont not have the bed perfectly flat at the edges is due to the uneaven transition from the metal plate to the plastic frame, thus lifting the edge of the magnetic sticker. You can fix that by shimming the plastic plate or cutting the magnet off a little.
Also the thing is, as there is a magnetic polymer layer ontop/the sticker, that is almost 2mm, this bed can NEVER be 100% flat, even if the bed sheet was 100% flat. Remeber that when comparing to a PRUSA or a voron that has magnets built inside the metal sheet or below the bed alike typical PRUSA, such a bed you can make more flat just by the nature that metal can be more flat than a polymer in production or by adjusting.
So you have a very nice bed.
But you have a nice results already!
Science lol
Awesome reply @Sticks , thank you 
Is this with the magnetic sticker on it?
If so it’s likely flat if you took it off like mine. The sticker or plastic tray that’s all around the metal definitely raises ehe edges.
It is the unmodified bed with all its layers and a cold plate plus spray mounted paper with the grid printed on (1-26, A-!).
I think some aluminium tape will do for now. I am to chicken to cut slithers of the magnetic sticker.
I seen it been mentioned before, but what about the bed mapping/meshing when the bed gets straight?
Now when my bed is “flat”, the bed leveling procedure sometimes only mesh around 5 points? I have not run orcaslicer, but is this configurable there?
Because i do feel that only 5 mesh points are a little bit to few. Any one has experience?
Same here. I suspect the software thinking it is “flat enough” to measure only 5 points. Good thing is that it significantly speeds up the process, bad thing is that in case the bed was warped in very specific way it would not see that warp and consider the bed flat.
Since I know the bed is OK, I don’t really care.
The way I observed it, it only checks 5 points if the temperature and plate are the same as the previous print. If it finds differences when checking the 5 points, all 36 points are checked again.