All of your rationales and logic make sense. However I want to add a bit on the part I quoted, sticking KNC value to ETH value does have its own downside, which is the correlation between KNC and ETH while KNC can be an independent economic.
Look at what Kyber is doing right now, everything has to contribute ETH as the fee, it means all trades (or in general, all service transactions) have to be evaluated in ETH, which is very hard and inefficient on-chain, the only way to make it possible is doing the calculation offchain just like Binance evaluates their trade with BNB value.
One point I think you are missing:
With burning model, we remove supply and distribute value to all of the holders, fair and square, people will have to look at it in a very longterm perspective to see the effect (like 10 years). With inflation model, there will be many more ways for people to access/make the tokens (preferably, by making use cases for Kyber) and if someone will just hold it for a bet in its value in the future, they apparently keep value for those new minted token, people can see its effect in a very short term perspective (like weeks). Being able to see short term outcome creates excitement and momentum for things to happen.