The distinction between “supinus… pedi­bus pansis …et pe­dum di­gi­­ti” for the circumference and “… a ped­ib­us imis” standing into the square, means beyond doubt that the extended feet must naturally exceed the square perimeter, even more than with the toes.
About a century after Leonardo, two skilled portrayals of the Vi­tru­vi­an ar­gu­ments were conveniently sep­a­rat­ed by M. Perrault de l'Ac­ademie Royalle des Sci­ences - Paris 1684.
His figures seem decidedly derived from su­per­im­po­si­tion on that of Le­on­ar­do (please click over), with the little variant of closer legs, in order to touch with the toes the base of the square – although not drawn and visible to­geth­er with the circle. But wasn't it rather a new com­pro­mise?: “hands and feet ex­tend­ed”  in and out from those perimeters?
A non-functional measure, however – if not to introduce the im­plau­si­bil­ity of the hypothesis of Vitruvius – since even here the way in which the feet are extended, apparently consistent with the roman text, does not appear to be containable within the same circle the­o­ret­i­cal­ly valid for the arms, reaching it with the half of the feet which pass it abundantly with their phalanges.

It is time to make technical clarity.

Let me resume the analysis from the basic plan, where the Leonardo's model looks re­stored into an exact square, defining the height and width of the stan­ding body, with an implemental premise.
Trying to resize together the square and the circle does not succeed 100%; minimal adjustments trigger meaningful changes to the navel po­si­tion, and a compromise is due. That's to say whatever change you do to the mod­el could put off target your exact measurements.
I experienced this issue, but now, viewing the best resizing to the max­i­mum extents of the circle and square and just to leave nothing to chan­ce, it comes difficult to argue what might have happened to that paper!