Peter, I agree with you that there is a big problem with high orphan rates, but this is not a symptom of GHOST, but rather a symptom of high transaction rates.
Whether you use GHOST or Longest-chain at high rates you must either resort to large blocks that propagate slowly or to high block creation rates. There is no getting around that. Both cause a high orphan rate (or perhaps the right term to use is high rate of off-chain blocks). we do not pretend GHOST solves this problem The only thing we claim is that GHOST makes the protocol secure at high rates -- the 50% attack can no longer be executed with less than 50% of the hash rate.
Going back to your original post:
We note that block propagation times are the primary obstacle for scalability.
The obstacle to scalability is keeping Bitcoin decentralized while scaling up; we know that Bitcoin can scale if we sacrifice decentralization - Visa and Mastercard are doing just fine. Ultimately you're proposing something that solves one problem - the high granularity of confirmations and the long wait associated with them - at the expense of scalability
with decentralization. So don't claim you've done anything other than presented an interesting academic analysis of a specific trade-off possible in the system.
This also suggests why the Bitcoin community has talked about the underlying idea in your paper repeatedly(1) among themselves, but no-one has ever bothered to investigate it more fully - it's obviously a bad trade-off in the context of proof-of-work. (with the possible exception of applying the idea to the p2pool share-chain)
h
1) I myself proposed it for my zookeyv key-value global consensus system proposal - #bitcoin-wizards 2013-05-31 - though in the context of proof-of-sacrifice.