But here a question of concepts arises, behind the numbers; and then, is it worth e.g. spending hours calculating numbers based on a fig­ure ge­o­met­ri­cal­ly inconsistent?

The drawing of the circle pointed out by Leonardo, plotted with a com­pass on the navel, as for the hands must be reduced from the point .e[xtrn] to the point i.[ntrn], thus losing its magic  tan­gen­tial­i­ty to the base of the squ­are; hence shortening even more the length of the legs.

But with any of such pre­ci­sion gap, no gear in the ma­chines designed by Leonardo Da Vinci could ever work.
The current adjustment from the reliable photo of the original (of Luc Viatour, www.lucnix.be. 2007-09-08), and even the use of bitmap tools may be somewhat critical; but the point is not how much the gauge be­tween the two rays is, but the fact that to join Leonardo's cir­cum­fer­ence, the arms should be rotated up to the intersection .a, not have to be for­ced within the square, a choice which does not comply any canon about.

Being aware that the "The shoul­der joint is ca­pa­ble of move­ment in ev­ery di­rec­tion, for­ward, back­ward, ab­duc­tion, ad­duc­tion, cir­cum­duc­tion, and ro­ta­tion", to ke­ep simple our case, after a long se­arch tho­ugh, I got a useful ref­er­ence and il­lus­tra­tions from the rich and ex­haus­tive "Normal mo­tions of the sho­ul­der joint" article, which shows how the complex sca­pu­lo-humeral joint ex­ten­sion is adducted along an arch when ra­is­ing the arm to one el­e­vat­ed po­si­tion point­ing to the sky.