The results from people who can be bothered to learn it's unique approach to 3D modelling are very good indeed, it's massively feature rich and stunningly good for a free application.
What bothers me with this take on the subject is not the time needed to learn, but the productivity once you learned, which is another matter entirely. Many modelling tools have already been revolutionary interface wise but very well planned (ZBrush for example is unique). To support a bit of everything doesn't necessarily mean getting the most out of every implemented feature (for example, you won't sculpt as good in Blender as you can in ZBrush or Mudbox, even though Blender is one of the few softwares to support proper sculpting).
Most Blender users dont want pretty pictures to click on, i prefer icons that actually tell you what they do, instead of some cryptic hieroglyph.
Then they should definitely checkout Newtek Lightwave. The interface is elegant, text written, key binding based, extremely modular and yet more productive than overbloated.
Featurewise Blender is very robust, however time is of the essence in the CG industry. As you can't be as fast a modeller in XSI against Lightwave or organize a scene in Lightwave in the same vein as XSI, with Blender you'll never reach the same results as fast as with leading modelling tools, and with the amount of hours you are likely to spend with any CG tool investing some hundred dollars for a more streamlined process IMO is a better trade off even for passionate hobbyists. It's certainly interesting as an introduction to graphics as it is more feature-packed than many commercial modelling tools, but now that modelling tools are not as costly as ten years ago a determined newbie should at least consider the idea of learning something which is both more proficient and applicable as a day job skill.
That said, the more the merrier, since it is still added research in the field and since it has some interesting features of its own.