For the same reason you don't build a bridge by grabbing some metal and stone, dragging it to the shore of a river, and start stacking it and bolting it together.
Trying just to build and run the Analytical Engine from incomplete plans is about as likely to work or be useful as typing in an incomplete computer program and "just running it."
But worse, since these are physical parts you're talking about, fabricating them without without sufficient planning is would inevitably lead to massive amounts of expensive rework as you find out that pieces you've built don't fit together and/or work properly because you didn't bother to check the design first. It's much cheaper, especially in this age of easy 3-D modelling, to build a computer model and test it before doing the real thing. (And by the way, it's unlikely that much of it could be built from 3D printed parts; the stresses many of the components have to deal with will require fabrication in metal.)
Since there's currently no plan in complete, ready-to-build form, that needs to be made first, and that's made by carefully studying all of Babbage's work over the decades he did it. Even then, we'll need to understand the plan and the techniques used in it, and that's also what we get through study. (Imagine you were given detailed plans for a modern bridge. Would you be able to supervise the construction workers to get it built safely and economically?)