Only agree with GregL partially. The original test does not consider the time spent to eventually append the rendered html to the page.
You should also check out http://jsperf.com/dom-vs-innerhtml-based-templating/259
This test will include appending the rendered html text to the page.
You can compare it with http://jsperf.com/dom-vs-innerhtml-based-templating/257
The original test case. You will find out that the difference across template engines is not that significant. You do need to consider more about the usability of a template engine, unless you are going to simply render a chuck of html text without put it to the client browser.
Besides, in my opinion, a good template engine should greatly help you maintain
the templates, help you write less to do more which is also the principle of jQuery.
doT.js is really great in performance, but I don't agree that it could help you code efficiently.
I am now working on an open source project to try to build a easy-to-use template engine. It can help you manage your templates in a tree-like structure, create template reference, define your own syntax, bind data to html dom.. please give it a try
http://jspopisno1.github.com/Beard
You can also have a read another question in stackoverflow
https://stackoverflow.com/questions/552934/what-javascript-templating-engine-do-you-recommend/8129492#8129492