No, it's not wrong and while the model used in film was about the size of your average Poser character, it used bump and displacement mapping taken from a model in the millions of polygons. So both a high and low res version was made (this is the case for most 3d these days - games or film).

The trick is, after you've completed your model or scene with a super high poly count, take a good long look at it. Try to see the things that you could have done with fewer polygons or fewer isoparms (if working with nurbs).  One of the things working with nurbs taught me is to capture a silhouette with as few isoparms (and I would later translate that idea to polys)  as possible. Once you have the volume and form of the silhouette down, detailing it becomes easier and will require less information. That's my take on it. See how low you can go, efficient meshes will lead to faster modeling for you and easier use for the end user.  If something can be done with a texture or displacement/normal map instead - do it. Save yourself time in anyway you can that doesn't sacrifice the quality of the work being done.

Currently, my method of modeling goes like this:

Create a base mesh (I have one biped base, one head base, and one creature base) in Silo.
Sculpt the base mesh into the model I want it to be using ZBrush or another sculpting app.
Take this high res model into Silo and trace over new, low-res  topology.

Working this way, I can concentrate on form and volume, while keeping the creative juices flowing the entire time. It's easy to lose interest if you're just working with polys or nurbs, trying to complete the mesh while keeping it as animation friendly as possible.  Working with a sculpting tool, you can finish your idea fairly quickly and then come back to it to get a better, more useable mesh from it. I've found many artists are starting to use this approach with fantastic results but the method isn't for everyone. Just because it works for me won't mean it'll work for you.

So there's not a real  right or wrong method for modeling in general but there is if you have a specific goal in mind or time table to fill. Animation requires light, clean meshes for anything good to come from it, while still images don't. Some jobs will allow you a very limited time to get a model done, so being efficient with your poly flow is important, while if you're modeling just for your personal hobby, you can take as long as you like.

Ultimately, it takes time and dedication no matter what approach you take or why you take it.  KillerBean, was/is a project being made by one man. At first, it was just a short done in his spare time. Now, he's putting the final touches on its full length feature. All one man. However, it took him years - more years than an average film (I think 8 or 10)- to complete and much personal sacrifice.  Is it Pixar quality? Not really. Is it good? Probably, if the short is any indication but that's the trade off you have to expect when working alone.

Anyway, I'm hoping this forum will become a place people can turn to for learning such things. I've already started recording myself working (with voice yet!) and plan to do some examples of my workflow. I hope it will spark others to do the same, since, everyone works differently. With a repository of different approaches and ideas to the same basic things, I expect it to be easier for people to feel comfortable with themselves as modelers once they see there's so many others with the same desire and all with different ideas of how to reach their goal.