It’s hard to disagree with point 1, yes it is new and as such still growing, you can see improvements on blockchain implementations on an almost weekly basis and you can start to improve it yourself if you have the technical skills. I wouldn’t call it flawed though, the principle is quite solid, it’s the implementations that may contain the flaws (not enough testing for one).
Point 2, I absolutely agree; this ties in with point 3 as well, if you need to use Oracles to push data onto the blockchain and you can only get the data through the oracles (trusted authority) then you’re not working decentralised, in that case blockchain may not be the right approach. Aside from that, security is determined by a number of factors, permissionless or permissioned blockchain to name a big differentiator. The risk in a permissionless blockchain (Bitcoin) is the accumulated power of a small group (the large miners) who can get a high degree of control on the success of the chain. If all Bitcoin professional miners move you will see a sharp decline in hash rate, subsequent use and ultimately value; while it is set up to self regulate, the adjustment of the algorithms is too slow at present to really address this (hey, one of the possible improvements to be done 
)
Point 4, there are various scalable solutions available now (think Tangle) that address this quite well, confidentiality is something that can be addressed in various ways (such as permissioned blockchain) where you can only see what you’re meant to see.
Point 5, agreed, they are currently certainly overhyped in terms of implementation, I don’t think it’s overhyped in terms of principle. If you look at what happened with Ethereum and Parity for example, there was a quite obvious exploit left in the code, this is a matter of implementation, not principle, which was ultimately exploited (malicious or not), leading to loads of problems. There is too much emphasis right now on getting new ideas out the door, rather than ensuring appropriate quality of the code and security aspects which would never have made it into a live environment in a more controlled software development lifecycle. Again, these are growing pains (very costly ones due to FOMO) but really just growing pains.
