perhaps you can tell me in what manner you think LN contributes to decentralization. The way I see it, it will inevitably end up a hub/spoke system - meaning more centralization. Also with node and path discovery being mediated by (other?) centralized actors.
I agree, although I think it's not the correct to say that the overall network topology will be be more centralised, just that one task the network performs will be divided into 2 layers (the miners on one layer finalising tx's, lightning hubs one layer below setting all the transactions up). And it's pretty much true that miners and exchanges could easily become the dominant players in lightning hubs, but it can't stay that way forever, because neither exchanges or miners will be around in large numbers 20 years from now.
tl;dr all Lightning does is split the transaction market up, not much point complaining that the same big Bitcoin companies will get all the business, because they do already now.
This would also take transaction fees away from the miners. I have nothing against the lighting network, even if it ends up being highly centralized. However it should not be regarded as a replacement for the every day transactions we do on the Bitcoin blockchain today. I also find your idea that the exchanges and miners will not be around in large numbers twenty years from now very strange. I suppose if we moved most of the transactions off chain that would be true. However if we do scale Bitcoin by increasing the blocksize I see the mining power of the network continue to grow as its value grows providing greater security for the entire network.
Today we are not reliant on third parties to transact on the Bitcoin blockchain directly and cheaply. It is not true to say that there would not be any difference under the model you propose, since in the model you propose there will be companies making money of transactions when before most of these fees where reserved for the miners, increasing overall network security, which arguably your proposed model does not.