Even this isn't enough: teachers suffer from small sample sizes and tests which attempt to reduce very complex realities to simple numbers. Imagine teaching 70 students, some of whom have significant problems outside your control (e.g. not showing up for class) - if that number changes by just a couple of students either way, your evaluation can go up or down completely due to random variation.

At this point, I'd also note that the people running the numbers tend to have very limited math skills – a great way to recoup some of the money we're pouring into standardized testing would be for each district to hire expert statisticians, scientists, etc. so the comparisons could be done by people who understand how to avoid measurement and analysis errors.