Thanks for replying, I'd be interested to see what you would come up with today using the same methodology, seeing as max single hard drive capacity has roughly doubled, global average internet bandwidth has increased 31% from 4.8Mbps to 6.3Mbps(sourced from Akamai State of the Internet reports 2014q4 and 2016q3), and we now have xThin and compact blocks to help significantly with block propagation time. Not to mention the usual improvements in CPUs(not that we're anywhere near a CPU bottleneck today anyway save for quadratic hashing when raising the blocksize, but I don't think that anyone would seriously suggest an increase without addressing that).
I don't think that the 17% yearly increase is too far off base considering current global trends(although I still don't particularly like the idea of centrally planning the limit, especially not that far into the future), but the 66% decrease first seems completely out of touch with reality.
I'd also like to point out to Luke that Satoshi envisioned most full nodes running in data centers in the white paper, not every single user needs to run a full node to use bitcoin. Not to present this as an argument from authority, but rather to remind us what the intention of the system was to be(p2p cash, not a settlement layer only afforded by the wealthiest and largest value transactions). That a lot of people want to continue to move in that direction shouldn't be a surprise.