I’m not sure if this is the place to report a possible error in the specification; if not, perhaps someone can direct me to the correct place.
Page 103 in the specification says this about orthographic camera parameters xmag and ymag:
The X and Y magnifications are simple scale factors, applied to the X and Y components of the orthographic viewport. As such, if your default orthographic viewport is [[-1,1],[-1,1]] as in OpenGL and DirectX, your COLLADA orthographic viewport becomes [[-xmag,xmag],[-ymag,ymag]]. This gives an orthographic width of xmag/2 and an orthographic height of ymag/2.
I believe the last sentence should say:
This gives an orthographic width of xmag*2 and an orthographic height of ymag*2.
That is, assuming that “orthographic width” is the same thing as viewport (frustum) width.