Not sure I agree with this. The article compares extremely low-fidelity wireframes with very high fidelity mockups. You don't have to choose between one or the other. Tools like Balsamiq allow you to create mockups that give a better approximation of the real thing with substantially less work.
I'm not saying you shouldn't create high-fidelity mockups, but get the usability problems solved first in a lower-cost environment. That could be as simple as a whiteboarding session or using index cards. You should be excited for a product/site not only based on how it looks, but what it does.
Mock ups (high fidelity) should actually precede wire frames and prototypes as a way of promoting a project in its very early stages to stakeholders. They are brainstorming outputs but don't really represent real design output yet, and must be promptly shelved after their internal promotional planning use.
The author is simply confusing interaction and visual design. The former involves wireframing, the latter almost always does not.
My point isn't about low-fi wireframes. (here's a slightly more complete version of the same wireframe: http://d.pr/i/N6Oo )
My point is that wireframes as a starting piece aren't always correct, and that if you do use them, it'll put you down a certain path, and that path isn't always correct.
I'm not saying you shouldn't create high-fidelity mockups, but get the usability problems solved first in a lower-cost environment. That could be as simple as a whiteboarding session or using index cards. You should be excited for a product/site not only based on how it looks, but what it does.