But "theoretical limits" have nothing to do with the cell packaging. Elon is usually very precise in the way he discusses technology. Theoretically any cell chemistry "limits" would apply to any cell size and shape.
The problem with 18650 is, to paraphrase Elon, that it wasn't designed to be the most optimal size. The limits *do indeed* have something to do with the overall efficiency.
Its size defines the packaging, and you can only cram so much mass or volume-efficiency in the cell if too much of the cell's mass or volume is related to its suboptimal case layout. If you assume that:
18650
18mm x pi x 65mm x 0.5mm cylinder wall = 1838mm^3, plus
( 18mm / 2 )^2 x pi x 0.5mm x 2 end plates = 254.5mm^3
total volume = (18mm/2)^2 x pi x 65mm = 16540.5mm^3
total usable cell volume = (18mm / 2)^2 x pi x 65mm - (wallvolume) = 14448mm^3
ratio = 14448/16540.5 = 87.3% of the cells total volume is available for active material
21-70
21mm x pi x 70mm x 0.5mm cylinder wall = 2309mm^3
(21mm/2)^2 x pi x 0.5mm x 2 end plates = 346.5mm^3
total volume = (21mm/2)^2 x pi x 70mm = 24245mm^3
total usable = 24245 - 2309 - 346.5 = 21589.5mm^3
ratio = 21589.5/24245 = 89% of the cells total volume is available for active material
Right there is almost a 2% gain in volumetric efficiency by simply changing the case size (if you assume my estimate of 0.5mm wall thickness is approximately accurate).
EDIT: Before someone says why not just keep making the cells bigger, there are other constraints on total pack efficiency. If you make the cells a larger diameter, you waste more of the overall pack volume on the air gaps between the cells because they don't fit together perfectly.
RE-EDIT: There is 1.49x the active material volume in the 21-70 as in the 18650. Theoretically, you should be able to use 66% of the number of cells to come out with the same capacity. Depending on how much of the cost is packaging of the cell vs active material cost, that could add up to big savings when spread over a pack's worth of cells.
They said 21-70 for MS-MX after the M3 ramp.
This much is true, but they also said we'd see X before we saw the D, and we all know how that turned out. Things like the order in which we expect to see things that are going to come about at approximately the same time, I don't put a lot of weight on.