A firmware revision to enable modifiers (eg. shift) on the mouse buttons is another great idea.
I would agree that one should be careful to treat statistics and direct experience separately. If the statistic help build intution, that's great, and if they differ, then the experience is probably even more valuable information.
I see what you mean about placing the spacebar at the periphery in order to maximize the stride for vowels.
Yea, is not comparing of single-button vs multi-button chords, but rather relates to whether dexterity is an issue for MCC.
Figure 7 shows that it is comparing chording using two thumbs, vs chording using fingers of a single hand. My understanding is that Lyons was saying that statically, those using two thumbs (very high dexterity) had a higher typing rate during the first hour or so (first four 20 min. sessions), but after three hours or more, those using a single hand had a faster typing rate.
There may be limits to the shapes you can easily form, but I personally compare it to guitar playing, where the shapes are less a concern as long as one uses a comfortable hand position, and learns to form those shapes with relaxed grip. With that, they all just become patters that with enough practice, 'muscle memory' can recall at will.
Even so, Lyons' results were for the default keymap, correct? The numbers will certainly differ for a keymap that has been optimized, as yours has, for a number of additional criteria.