Have you looked into profiling with strace? You should be able to see the actual system calls each of your tests are making. This will probably give you some good insights into what accounts for the speed differences.
I think to be one of the principle developers at a start-up, you need to do better than feeling like your "skills are lesser than a huge community of people working together".
IMHO, the two/three primary developers at a start-up should be able to code up a new framework that better meets their needs (though possibly less total features) in less time and with fewer bugs than a huge community. It shouldn't even be a close race. Also, by developing their own framework they will know exactly what parts suck and still need work and be better able to support it at launch.
"Why the hell would you write your own framework and reproduce all the features of an existing framework"
For fun.
To learn a new language.
Because you have a new idea to explore.
Because sometimes people do invent better wheels (I love the irony in this phrase, given this: http://en.wikipedia.org/wiki/Mecanum_wheel )
So you are saying for a large project it might be justifiable to implement your own framework?
I would think for a smaller project you could quickly put something together in less time that the learning curve for some of the frameworks.
Bug-filled/half implemented seems to be more a comment on programming ability than the decision to write a framework.
Fortunately for many people, others have decided they would rather write a new framework than use something that exists.