Also, creatures that don't need to breathe, eat, drink, etc could survive in the walls if they teleported there
I imagine their supple (in)organic flesh might have a touch of trouble co-existing with solid granite.
However, your point about scouting reminds me of an interesting thought I had. In a recent fortress of mine, I was ambushed by elves in my second spring. As this was an outdoor town, I had absolutely no defense. So I drafted all seven of my dwarves and sent them at the enemy, driving them off with only one casualty, if memory serves. Just for fun, though, I copied the save to see what would happen if I left them as civillians. Unsurprisingly, I got my ass kicked, but only lost three dwarves before the elves called it a day and went home. Only one of their twelve or so ambushers had died. This begged the question of why they even bothered attacking in the first place, and inspired the train of thought I'm about to share with you.
In a game as complex as DF, we should not be looking at balance, but rather at motives. Balance is desirable, yes, but balance for its own sake is bound to feel forced and arbitrary. Instead, conditions in the world should encourage behaviour from both the player and AI to gravitate towards a balanced solution, Le Chatelier-style. But I'm getting ahead of myself.
Right now, ambushes and sieges only really differ in size and stealth. Both simply seek to kill as many dwarves as possible before suffering enough casualties to want to retreat. Ideally, they would act differently based on their goals.
Ambushes, being stealthy, hit-and-run squads, would have short-term objectives. If a site is poorly defended, kill as many inhabitants as possible and steal whatever's not bolted down. If it is more heavily guarded, attempt to scout out its defenses while disrupting the outdoor workforce if possible. Either way, retreat at the first sign of superior opposition.
Sieges, on the other hand, should have goals related to the long-term welfare of your fortress. In particular, they should want to claim the fortress for themselves, but since that's likely impossible, they will want to put pressure on the inhabitants and keep them from leaving.
This leads to the player's motives, or rather the lack thereof. Under the current farming and brewing mechanics, there's hardly any need to ever leave the security of your fortress, except perhaps for wood; therefore, a siege that keeps you in is accomplishing very little. Posts in thisvarious threads have suggested ventilation and running water as potential weak points, among other things, and these are valid ideas. But think beyond the fortress gates: A siege is preventing you from leaving the
map. The Army Arc will lets us do battle in foreign grounds at our leisure, but it might not be long before we have a strong motive to send armies out for our own good. I refer, of course, to the County Arc.
Let's say you're a barony capital, which most fortresses are around the time sieges start. Right now it's just another set of annoying mandates and that goddamn economy. In the future, however, it may entail receiving tribute from the surrounding countryside in exchange for the protection of your armies. Sieges, therefore, would trap you in the capital. If you don't meet the enemy in the field and break the siege, not only do you lose your income, but you risk being demoted, with all the mood penalties that may bring about. Disgruntled ex-baron, anyone?
I'm not saying that sieges shouldn't have ways of crossing your trenches, busting your walls, and disarming your traps. But an elegant solution is much more satisfying then a silver bullet, no matter how awesome that bullet may be.
tl;dr version: The County Arc will make people leave the security of their fortresses.