I generally prefer dungeons that make sense. For example, when I build dungeons, I often look up old blueprints of castles or mansions and base the dungeon's layout off of them; none of this "labyrinth" nonsense where corridors and room placement makes no sense. It also helps when determining the ecology of dungeon; where is the kitchen, food storage, and commodes (all potential sources of food for its denizens), where are the bedrooms (smaller nests for vermin) and living spaces (dens for larger predators), what walls have structural damage or where are the balconies/outer doors (how did the dungeon's habitants come to reside there).
I've only ever really done one, and not really a dungeon (the remit being "Oblivion's castles don't have many NPCs; why don't you add some?" So I got a bit carried away and massively extended them, including caves, new districts of the town and so on. And then ran out of energy, unsurprisingly, so it was never finished).
Er anyway. So as much as I
tried to make it actually make sense, in that it fitted together logically and contained no physical impossibilities, I think I would've actually done a better job if I'd taken your approach and actually researched it properly. One glaring example is that actual "contemporary" architecture tended to not feature long corridors: the odd gallery maybe but they tended to be a series of rooms as a corridor is essentially wasted space and a bit of a modern invention. But I liked them, so that's the way it happened.
Of course I would've been even better advised to just follow the bloody remit. Maybe that way I would've finished it... though it's not like I have an especially impressive legacy of ever releasing my mods, finished or not.
