@Eyeorderchaos: I don't know any other apps, but I'll defend Poser's setup on
biomechanical grounds.  Not 'external' biomechanics, ie the biomechanics of 
a cadaver, but the biomechanics of the living internal system.

When we make a move, our brain doesn't pay any attention to the limits of
the joints.  Instead, it  (1) defines a goal; (2) tells a magnificent set of position
and strain and acceleration sensors embedded in muscles and tendons
what they should be sensing on the way to that goal; (3) receives an error
signal [negative feedback] from the sensors, which decreases as they approach
the desired goal; (4) controls muscles to minimize the error signal along
the way.  

For an extreme but illustrative example, think of Ray Charles at the piano.
He can't see the keys, and because he never saw how other people move,
his body is thrashing around with the music.  Nevertheless, he always hits
the right keys.   Now if his brain were giving commands one joint at a time
(To reach G#2, set right biceps to 43.5 degrees, pronate radius and ulna
to 32 degrees, etc) his hands would rarely even approach the keyboard,
let alone the right key ... or else he would have to wear a full-body cast
while playing!  But because his brain keeps perfect dynamic track of all 
the positions and accelerations between chest and fingertip, his body is 
free to be the music.    [Yes, I know Ray is dead, but he's still a wonderful
example as well as a wonderful musician.]

The range of motion of the joints is more a result of the sensory system
than an initial limit.  When the system goes wrong (such as athetoid CP
or dystonia, where the sensors give positive feedback part of the time, 
leading to undamped or increasing oscillations) the muscles are perfectly 
willing to overcome the limits of the bones and ligaments, leading to 
subluxated joints and even broken bones.  

Poser's IK system, though it misses the accelerations, does a pretty good 
job of matching the internally driven biomechanics.  You put the hand on
a goal, and the joints come into proper alignment through [what seems to be]
an internal set of equations that reproduce the real negative feedback.

My python page
My ShareCG freebies