This is a strawman. While a fully objective evaluation is hard or impossible in most cases, it doesn't mean that it can't be somewhat objective. The extremes are pretty easy to identify - amazing performers who implement lots of features with few bugs and who are easy to work with vs the asshole who gets nothing done. The difficulty is in the middle, and while subjectivity is unavoidable, you can use objective metrics as data points.

Also, a well built subjective review process is better than no process, for instance, using peers and customer opinions as input.