* Stop using Devel::DProf. It has been replaced by Devel::NYTProf. You'll get much better information.
* Run your page requests 10-100 times in one profiling session, or wait to enable the profiler until after the first request. If you don't, the setup time will swallow everything else.
* Profile a real application, or at least something with a database request in it. If you profile a "hello world" app you'll wind up spending hours working on things that increase performance by 0.1%.
* I think it would be a mistake to use HTML::Template::Pro. Its only real speed advantage is when running as CGI, and if you care about performance, you should be using mod_perl or FastCGI. There are faster templating tools than HTML::Template, but I've never seen it be the bottleneck in any real application.
Actually I would say security is a slightly higher concern than performance. For the forseeable fortune this is likely to be a VPS running a lot of sadly obscure websites. As such I am not using fastcgi or mod_perl but rather suExec.
As such the module load startup is significant. I think that also shows why I stop using Titanium until I am using all the modules it loads. Still there is no reason to completely ignore performance or deny speed ups where available. Also the script does have a database backend though I will keeping rerunning the profile as it gets bigger. I am not laying to much by this experiment but it seems I have to agree about HTML::Template::Pro its contribution to the speed up was quite small.