I read your post, and I've re-watched some of the videos you posted (most of which I had already seen before). Oddly, the validation is always conveniently overlooked. Yet, Baird does an excellent job at repeating "Fast, Secure, and Fair" to make sure the terms are well-ingrained into the viewer's subliminal and subconscious mind. He even starts out by saying it's currently meant for permissioned blockchains, yet he repeatedly compares it directly to other blockchains, again repeating the mantra, in comparison, that Hashgraph is "Fast, Secure, and Fair".
I'm not saying that Hashgraph doesn't have its particular niche, but the performance claims, by Baird included, are frankly only achievable in very particularly tight usage cases. If I'm still missing something here, perhaps @dan could fill in any gaps in my understanding here, as I'm quite sure he understands this far better than most, myself included.
And once more, for good measure, to really ingrain it into your minds...
You say that DAGs aren't Hashgraph. However, do you really believe that if graphene swapped out its currently foundation in lieu of Hashgraph that it would really be orders of magnitude faster? That seems quite unlikely to me. But perhaps I will be proven wrong in the near future.
As such, I do look forward to seeing some of these performance claims validated in a real-world application such as STEEMIT in the near future (quoting from your post):
Hashgraph is both an algorithm and data structure that doesn't use blocks or wait for blocks to be confirmed. There are no blocks of transactions, but instead streams of events that can be segregated into clear transactions. Since there is no waiting for blocks, Hashgraph can do hundreds of thousands of transactions per second. In comparison, Bitcoin can do 7 transactions per second and Ethereum 15 transactions per second, but even these rates are at their best performance and not their worst.