There's no difference between an activation threshold of 75% (or 51%) and 95% because pro-fork miners can just start rejecting blocks from anti-fork miners to increase the percentage. Miner voting is a stupid concept for hardforks -- it's the economy that matters in a hardfork, not miners.
You are of course correct in that 51% of miners can control all the miner voting. Implementing a hardfork is a difficult process and a variety of things need to be considered. Miner voting should not be considered in isolation.
With respect to a one off shift to 2MB, I think there is broad support from users, businesses and miners for this kind of compromise proposal. However given the controversy, particularity the number of developers who do not want to go down this path, there is elevated risk with this hardfork. Therefore I think it is prudent to have a mining activation threshold of 95% or more, in order for me to support the hardfork to 2MB. If miners start orphaning blocks that do not vote their way, many will consider this a serious attack and withdraw support for the hardfork.
However, according to Brian Armstrong the threshold is "only at 70% for classic".
Source:
https://twitter.com/brian_armstrong/status/688428800959332352I consider 70% dangerously low and demonstrates a lack of confidence in consensus by Classic proponents. It is therefore important for users and node operators to continue to rally behind Core, to prevent a potentially dangerous and catastrophic loss of consensus. If Classic adopts a 95% activation threshold, then I may switch to supporting Classic.