The fee here is not the problem. I have benchmarked ghash.io against btcguild and eligius with equal hash rate over a period of several months. BTCGuild came out on top, and at the time it had a 3% fee and the other two were 0% fee. If there is one thing that will bring ghash users over it is more frequent blocks. Keep in mind that most people mining there are also using cex.io, and these are mostly people who think that 0.005 BTC and > 20% fees is a good price for 1GH/s. Not the brightest people.
So we have a chicken and egg problem here, but I think the only way the fee could be safely removed is if the PPS split is removed and DGM is taken off the sliding reward schedule similar to the way EclipseMC runs theirs, where every block is paid 25 BTC + fees regardless of how long or short the block is. A donation mechanism would then have to be used to recover the costs of running the pool. Then finally after all that work we'd probably still get no one from ghash.io moving over.
I am not sure where you got most of your info, but your very incorrect. cex.io had a public statement where they said that 25% of the hashrate they have is their own. That means 75% of the hashrate is by miners just are pointing there local miners at the pool and not with cex.io.
As for a 0 fee, I am not saying a 0 fee will make us more money over the long term tho I think it will, having a 0 fee encourages miners to switch as they see our pool says no fee then they switch.
Anyone running a 0 fee pool should be suspicious. There is an ulterior motive somewhere, and understanding that motive would be required before any reasonable person would switch. Again, I'm not saying people are reasonable, but I'd have even less faith in mmpool if it went to 0 fee.
If you believe that only 25% of ghash is cex then more power to you. You might be the only one who believes that.