Fake bitcoin classic nodes would not be able to raise the percent of support since they aren't actual miners. At the end the amount of hashrates are mattering.
Hashers can throw their power behind a classic miner then later withdraw their hashing power.
And when a fork happens then when support for this new system raises a high value which means everyone will switch then since it is very unlikely that they will drop again then.
As I've stated I don't think we are just dealing with sensible people making sensible decisions.
If Bitcoin Classic were to stick to the way that soft forks have so been done (95% before activation of the new feature) then all of this controversy would go away.
If the vast majority of people do want 2MB blocks then I don't see that 95% should be a problem (and for the record I don't really care if the blocks are 1MB or 2MB).