No, you are totally right. The way you describe is how an actual A/B test should be done. But in practice companies resort to making changes in A/B test which involve changing more than one variable (in this case button style and text). That's because they can't afford the time and traffic it will take to methodically test one factor at a time. Also, primary motivation for many companies is to increase conversions. Knowing and deducing what increased conversions is a secondary priority.

By the way, I did not do this A/B test. I just interviewed one of my customers who did this test.

Ah, now I get it. Sorry, for some reason, I assumed you did the test for the customer.

I shouldn't doubt your A/B testing ninja skills.

The button test, even combining style and text, is still quite a small change to test. Don't your customers want to test more drastic changes?

Yes, they do. See our case study library (we have more than 15 A/B tests): http://visualwebsiteoptimizer.com/case-studies.php