Doesn't it work to just draw all the geometry first,
then push an identity matrix and draw the text after?

Here's a complete example that shows drawing text on both
the object and a fixed screen location:
http://seriss.com/people/erco/fltk/#OpenGlTextOn3D