There is no way to tell exactly how much of the TX volume LN would acquire in a week, month, in a year. On-chain scalability should not be based on speculative patterns but rather past patterns (e.g. once there is a history of LN usage, then factor that in as well).
Yes. But you're not disagreeing with me, that's what I said too: wait until Lightning has some reliable medium term stats before we start deciding that 1MB, 2MB or anything else is needed. How can you know before Lightning is established?
Besides, LN is not a close reality yet. It's currently being tested, and it doesn't even have a GUI (from what I've seen). It's far from being part of the main-net and even further away from being consumer friendly.
Further away than a successful (planned) hard-fork? Not sure about that.