Stored pose: the only thing that was done with verticies is translation by apropriate joint position. Type2 joints: shoulders, hips. Notice that hip-knee(belonging to hip) vertices orinetation is vertical, but knee-toes (knee, ankle, toes) are horizontal (Same goes for arms). Why? We get to this later.
Bind pose: we use type2 matrices to transform children positions and their vertexes. ================================================================ What does this all mean? Animation is stored in such a way that transformation required to bring arms, legs to proper position is stored in their respective parent type2 joints animation track. This reduces interdependance of euler parameters(gives more degrees of freedom so to speak). It does cause a slight complication to the type2 joints themselves: In order to move properly this embedded rotation of children, that we talk of in previous paragraph, must be compensated. In the simplest case this is done by just multiplying matrix produced from type2 joint's animation track by the inverted matrix of type2 joint.