Can you substantiate these claims? It sounds a bit as if you're generalizing based on a bad experience in trying to use Drupal. And to be fair, Drupal does take a while to get used to -- you won't build an awesome site in the first week after untarring the package on your server.
However, all of the following sites were built on Drupal -- and there is nothing intrinsically fixed about their designs or their ability to scale, etc.:
-The Onion: http://www.theonion.com/
-MTV UK: http://mtv.co.uk
-Yahoo Research: http://research.yahoo.com/
-Ozzy Osbourne: http://www.ozzy.com/
-Moby: http://moby.com/
My closing statements were definitely hyperbole, but I fundamentally stand by them, both as a web designer and a developer. My opinion comes from experience, which I believe is more informed than looking at a bunch of sites without knowing what went in to each one.
The thing with Drupal is that it caters to generic needs very well. It will get you 95% of the way on any project with moderately standard functionality. However the last 5% includes fine design and usability touches that I take as a point of professional pride.
I had one project where I built the whole thing in Drupal in about 20 hours. After crossing the finish line, the client was unhappy with the checkout flow from the e-commerce section. I wasn't happy with it either, but it was just the way the module worked. Although there was no reason we couldn't re-work the e-commerce module, there was no modularity or unit tests to help with this kind of refactoring. I estimated the rewrite to be somewhere in the 20-80 hour range given the potential for massive breakage. In the end we decided to cut our losses and spend 40 hours rewriting in Rails. We fixed a lot of other design details along the way, and we have a lean 2000-line code base that we could extend in any reasonable fashion. Sure we couldn't drop in a forum/gallery/blog/library at the drop of a hat, but is it more important to have a million pieces of generic functionality at the tip of your fingers or to get the core user experience correct? It's bad enough when you have to fight with the client to get the UI right, but what I find truly unbearable is when everyone's on the same page, the ideal UI is clear, the logic is simple, yet you're encumbered by technology that's designed to solve 1000 extra problems you don't have at all.
They aren't groundbreaking websites, however. They do their job well and all of those sites are successful because their owners/content was successful before their current drupal incarnations were developed. I'm also aware that there are some sites that have launched their 1.0's as drupal sites, I think it's pretty safe to say that these sites are familiar conceptually (like digg, like slashdot, like etc.) In that sense they don't have "amazing" designs.
Perhaps we should rephrase this as "you will never create an innovative desgin in Drupal." I'm generally in favor of Drupal, but I think this is probably the truth.
It's also not a bad thing. Websites need to excel at presenting information above all else, and drupal sites provide a lot of power to make this happen. Most websites are adjuncts to other things that organizations and individuals are doing, and they don't need to "shift the paradigm of web design and architecture." Lots of people need websites like this, and drupal's ideal for these cases.
---
I think the "you'll never create a successful startup in Drupal," statement requires a bit of a clarification, but it's probably mostly true: given the constraints and assumptions of the platform, you're unlikely to create a new kind of website in Drupal that will succeed on the Internet. The next boom of new websites (like blogs in 2000, or social networking sites in 2004, or webapps in 2008) isn't going to be built in in Drupal first.
But it will be built in drupal second (maybe,) Drupal is great for making custom instances of websites that are targeted at niche/specific audiences. And that's a huge chunk of websites. It doesn't make a lot of sense to build highly local sites like "digg-chicago" and "digg-Sao.Paulo" all in custom-built perl or ruby based sites, when something like Drupal promises to make development and maintenance much more simple in the long run. These kinds of sites have a great deal of worth and potential.
---
As for lean and scalable.
Totally the truth. Drupal can scale, but it's a pain in the ass. Which means, a sliver of the huge gains made in development times and maintenance times is absorbed by increased architecting costs. Lots of memcache and hardware can fix this problem. But if you get to the point where you get to that level of demand, presumably you'd have to have a sysadmin on staff anyway, so I'll concede the point, but I think it may really be a draw.
I share your dislike for Drupal, but I wouldn't go so far as to say it's unimpressive. My gut instinct is that Drupal is about as good a hybrid framework/CMS as you can get. That's not to say other approaches might not hit your or my sweet spots better, but fundamentally I think it's operating at the wrong level of abstraction.
The problem is to give you anything that works out of the box a CMS has to make a ton of assumptions. Drupal goes to great lengths to make everything configurable and hookable. It's hard to come up with a use-case that can't be coded using Drupal hooks. The problem is you face the crushing weight of a system designed to meet the needs of any and all potential websites. Even though deep knowledge of the system allows you to do amazing things with very few lines of code, the system itself is a straitjacket where the cost of seemingly trivial customizations can be crippling. If you find yourself with a scalability problem that can't be solved with naive caching you may well be totally fucked.
Contrast to web frameworks like Rails. The goal here is to give you a set of raw tools to make it easier to handle the trivial yet repetitive tasks that come up over and over again in web development. Unused features in Rails don't introduce cognitive load on the developer. If you've spent any time writing CGI scripts before then you can see the reason for almost every feature and design decision.
It's possible to build a career on top of Drupal. Becoming expert in it will allow you to "solve" a wider range of client problems faster than any other CMS or Framework. The problem is that you have to embrace compromise at all levels. You will never create an amazing design in Drupal. You will never create a successful startup in Drupal. You will never write a lean and scalable site in Drupal. If you are willing to compromise you will find a huge swath of potential clients who think $1000 is a lot to pay for an e-commerce site or want a do-it-yourself solution with a little hand-holding.