Sweaty, the principle you're observing is a normal and predictable feature of human anatomy. From
Dynamic Figure Drawing by Burne Hogarth, p. 136 [bracketed information is my clarification; italics and standard parentheses () are original to Hogarth]:
The Joint as Pivot; the Member as a Radius
Let us point up the significance of the concept which emerges from the foregoing demonstrations [the depiction of circles and ellipses in perspective]. In the perspective circle, we see that the variable lengths of line are identical and equal in different directions in depth. This means that while these lengths appear to change -- because they are put within the control of an ellipse -- they are actually unchanged in length.
Should this proposition be applied to a pivoting member of the body -- to an arm or a leg -- then the over-all control of such members would be maximum, no matter what their direction in foreshortened space, Indeed, we should then have precise control of the irrational lengths of forms in space. Perhaps the following will serve to illustrate this concept to the reader.
[Insert: An ink drawing of a nude man in profile, with three right arms arranged in different poses and three left arms arranged in different poses, with a semi-circle drawn in dashed lines through each of the elbow and wrist joints, indicating the proper and correct placement of each along a circular arc. This dashed-line circle continues through the torso to bisect the navel. This bisection point is marked "A".]
Suppose we look at a simple side view of the upper body and study the movement of the arm. First, the length of the upper arm (extended left arm) will be given the correct proportions by establishing the elbow, taken at the inner bony projection (the condyle), on a line held equal to the position of the navel. This position occurs midway between the rib base and the pelvic crest, carried across the mid-axial waistline of the body (A). Opposite, we see the right arm also taking this equivalent position at the elbow (A). Now, using the shoulder as a pivot, the extended arm swings up. The swing describes two arcs -- a short radius, and a longer radius -- held by the limits of the upper arm-to-elbow and the lower arm-to-wrist (see dotted curves). To the right, the elbow rises backward, describing an arc, using the shoulder as a pivot.
It is obvious that, in this multiple sequence of positions, if the arm length (on each side) is given to the arc of the circle, any position on the arc must be held correct as a radius. Hence, if the rotating elbow is drawn on such an arc, it, too, must be held correct at any position and can be developed fully as an arm, without distortion.
The author goes on, over the next seven pages, to describe how this arc circle also controls the pivot points of the leg, and can be tilted into foreshortened perspective using an ellipse.
I hope this clarifies the matter for you, Sweaty, and puts to rest your stated opinion that limb lengths extend horizontally ad infinitum rather than pivoting along an arc circle.