Well, the main reason is modularity:

The purposes of modularity are: reuse, and separation of concerns.

The question to ask yourself is can either those modules, who's sole purpose is to initialise instances of this particular class, ever be used without the class module?

Based on your description, the answer is no. That means those one function libraries are "close coupled" to their parent. They can have no purpose without their parent; nor the parent without (at least one) of them.

There is another acceptable use of modularity. That of breaking up large single files into manageable bits. Generally, this practice shouldn't be needed, because components naturally break down into manageable sized-files, but sometimes it makes sense to introduce artificial granularity.

But in your case, without knowing which XML/JSON readers you are using, it is hard to see that populating a hash from a file in one or the other of those formats could take much more than 10 or 20 lines each. Separating out of 40 lines from the main file, no matter how big it is, will make no difference at all to its manageability.

At the extreme, almost every individual file becomes one line of executable code. And the few that aren't then become meta code dealing with how to manage those one-function/one-line files. You end up with 100 lines of executable code, spread across 50 files. 50 of those lines are single string compares, and the other 50 are a bunch of hideously, slow string evals, who's only purpose is to load the the 50 string compares.

I'm not suggesting that you'd go that far, but do think carefully before arbitrarily breaking out lumps of code into separate files and name-spaces.

Anyway, you asked the question, I've added my perspective to the pile of replies. Now you can choose your favourite, or ignore them all :)