I agree with those who believe there is probably some "tolereance drift" both within a starship and across a starship's class. Since we know starships are not built as perfect copies by replicators, there is likely some variances in the tolerances of each ship owing to a whole constellation (excuse the pun) of factors. And heck, what is not to say there is not some "tolerance drift" in a replicator?
Yes, it's 300 years in the future and manufacturing techniques will no doubt improve steadily over that time, but these machines are being built partially on the ground (in a gravity well) and partially in space (in a microgravity well). Such variations could very well impart their own variations on the structures.
Now, chances are the tolerances for a structure as large as the mated surface areas of the GCS saucer and neck are tight enough to allow interchangeability of "stock" saucers (ones that were built to the same specification and as such likely have pretty tight tolerances). Anything within a couple millimeters (which strikes me as reasonable for mid-24th century tolerances) would likely be fine.
I imagine the docking clamp system itself forces the two pieces into the proper alignment as they get closer (since I doubt the ship's motion and position control system is accurate to the nanometer level), so provided the tolerances are close enough, things like EPS taps and turbolifts should all mate correctly.
Well, the only way you could be sure of something like that would be to design in a "flexible" interface... so that the elements in the mating section would be adjustable on-the-fly. Thing is, that makes for an even MORE complicated assembly, doesn't it.
Is it theoretically possible for a "docking system" to be designed which would be loose enough, or "adjustable" enough, to permit every saucer and every secondary hull in the fleet to connect? Oh, I'm sure it would be.
But the problem then is that you don't have an OPTIMIZED solution for ANY combination... the interface would be weaker (probably by a significant margin!) than it might otherwise be. Yes, you could have the turbolift junctions designed to allow a transfer without having to individually tweak the shafts to line up within, say, .005"... and allowing as much as 0.5" offset between shaft centerlines instead. But to do that, you'd have to have every car slow down and "crawl" through the interface as a result.
There are COSTS to every design decision you make. SO... you have to decide, will you build in the capability to mate every saucer to every secondary hull... or will you optimize the performance of a specific combination of components to perform the best that they can, TOGETHER?
(Those are directly OPPOSING perspectives. I know someone is going to say, now... "but it's the FUTURE so that doesn't count anymore." Bullshit... that argument only holds if you believe that this "future" will be set in the Land of the Lollypop King, with marshmallow skies and Koolaid Rivers...)
So... YOU are in charge of starship production, and you have to make that choice... which do you choose?