Let me see if I can help. So... what I think you're saying is that this is a hard task to estimate because to do the estimate you need to actually do the work, i.e. you need to fully understand the code. Is that correct? ( I know, no more questions.
)
But it's like this with any estimate. You just have to give the best estimate you can. Your manager will understand if you are a little late, as long as you don't have consistently short estimates and are always late. Most of the time you will do a little of the work to get a good estimate.
If I were tackling this problem, I would first try to get a very high level understanding of the code; not even go through each class, just get a basic understanding of what the program or API does. Then I would take one of the classes and actually do the work. Don't take the hardest one or the easiest, but somewhere in between. You should be able to get a feel for how long it will take to do all of them based on commenting the one class.
Are there unit tests? This can be very valuable in determining how a class was meant to be used.
Also, I assume the person that wrote the code is no longer there? I'm probably pointing out the obvious, but that person would be the best resource and could affect your estimate.
There must be some type of projects in particular that would be best suited for Clojure. I get that you can have OO concepts in Clojure, but it sounds like the idea is to solve problems in a different way using all functional programming? I just can't imagine modeling the world in software without OO concepts.