Hmm. It seems, I am still unclear with my complaints. Sorry for that. I am Russian, and even on Russian I have some problems with explaining my thoughts. Let me try it again, sorting my complains.
1. Yes, I know that "minimap" in the corner actually is graph. I simply see no reason to be it a
such graph. Why to? It can be a "real" map too! Actually, most of dungeon maps is the same graphs.
2. "One-way" graph by itself is too artificial. There isn't any buildings, except simplest ones, that haven't real loops, double staircases, secret (unused) doors, several entrances/exits etc. But... yes, it can wait. It could be realized in long-time perspective - or not. It would be a pity, but I can live with it.
But I see no reason for it to be one-way graph in
such, "vertical" design. Yes, current graph logically is equal to the same graph but with branches going up and down. But for human perception it is still different!
If it was "flat", it could be bearable - but I see ladders here and there! And all of them goes only up!
3. Well, let all this buildings to be one-way only. But then only difference of current graph from "real" map is this "autosorting" of all things. But... is this necessary? Really? For me, it only create needless level of artificialness. That, in turn, breaks suspension of disbelief somewhat. And, well, creates some confusion too. Also it gives totally out-of-place possibility to get to maximum floor simply selecting rightmost exit. Precognition, I think. Is there
another reason to have "sorted" map?
Enough for now, I think.
[edit]
Oh, wait, I just recall last "issue".
4. All building maps are "collapsing", adding up to mentioned "pyramid" design. Room of 6 cells can lead only to room of 6<=cells and have connection only to max. 6 rooms. Every one (if it's more than one) will be at max 5 cells, etc. In the end, 1-cell room can only expand to the same room or end the chain.
I think, it's an example of design choice that affects gameplay and results in predicable (so artificial) results. Each "base" level gives only size that <= it. And I don't think that it is needed, actually. Of course, you still need this "actual" width, but it not needed be "visual" one!
Let's try already described design:
________
_______|_8_|10#|
_______|_6*|___7#__|_9_|
|_4_|_3*|_5_|____2a_____|
|___________2__________|
|___________1__________|
Why it must be layed out such way? It could be made like this:
________
___|__8_|10#|
_______|_6*|_7#_|_9_|
|_4_|_3*|_5_|_2a_____|
|_______2___ ___|
|_1_|
Such way, of course, not very "nice", comparing with previous variation. But it's more "natural", allows more variations and not so predicable! Also, for niceness it could (and actually must) be "filled" with "empty" blocks:
________
___|__8_|10#|
_______|_6*|_7#_|_9_|
|_4_|_3*|_5_|_2a_____|
|_______2___ ___|__0_|
|____0____|_1_|_0____|
Of course, logically this scheme don't differs from previous one. It is based on it after all.
But a) it looks more natural (for me), and b) it allows "expanding" that not possible in current design. Like that, for example:
________
|9_|10|11|
|__8____|
|6|7|_0__|
|3__|4_|5|
| __2____|
|1|__0___|
Now it's really all that I can say. For now.