This experiment is not very rigorous. The Q4 2011 change in the proportion of UI bugs could be caused by the addition of back-end code with little UI (such as maintenance tasks), or by an upgrade to the bug tracking software that encouraged grouping multiple UI bugs into one bug report. And the author didn’t calculate whether TDD slowed down their work – it is not clear that TDD is better if it avoids defects but also increases development time.
You can find more rigorous studies on whether TDD actually works on Google Scholar: http://scholar.google.com/scholar?q=test-driven+development+...