Giving less of my info to sellers is therefore less a technological characteristic of a payment system, and more a choice by the system's creators. There's no technological hurdle to even having a credit card which gives a bit less info. Theoretically, the credit company only needs to see my expense, current available funds and some transaction/merchant identifier, not exactly what I spend it on or even who I am.
I'm hoping some of the new NFC-based payment systems will make privacy a differentiator. They'll need something if they want to get into the market with such established players. We'll see what Apple comes along with. I've been slagging them quite a bit, but they do try to make privacy a selling point. Perhaps here too?
As for the second issue, I am sure that a form of chargeback could be implemented on top of crypto. Your first link just shows one way to do it - but doesn't link to any implementations of this. The second link explicitly mentions 'no chargebacks', so again, it's theoretical in the sense of not existing.
The hurdle here is again not technological, but the apparently inability of the crypto community to think that people might want this behaviour. Sometimes even sellers. Where I live, allowing return of unused goods before a certain period of time has passed is mandated by law. Payment methods which does not have chargeback may actually complicate seller's business in this case, since than they have to be even more careful against internal fraud and pay an extra fee for the return as if it were a new transfer (if not in cash).
Thirdly, fraud and ransomware are getting worse and worse. I'm sure eventually there'll a crackdown. I just can't predict how, where and when. But crypto would be a very convenient target politically; this doesn't yet require targeting any large stakeholder.
Alas, the one thing the crypto community is right about, is that many economists are just waiting to be able to implement negative interest rates - but I doubt crypto would be able to change anything there.
Ultimately it boils down to, without comment on the desirability of making it possible to purchase illegal goods, can your proposals match the low amount of information darknet markets require along with the low amount of friction in making a payment? (A shipping address for physical items is the maximal information needed by any party in that whole system.)
For your second point, you can opt-in. Basic escrow has existed since the beginning with m-of-n signature transactions, various wallets have UIs for it. I'll just leave this here: https://en.bitcoin.it/wiki/Multisignature For "risk-absorbing systems" for merchants, those too have been around. BitPay is probably still the most popular, and it's several years old now. https://bitpay.com/ If you as a merchant use them, you can if you want not even bother dealing with BTC and its inherent risks at all but still accept it as a payment method.
For your third point I agree (elder)¹ fraud is a problem, and that if the scammers extract payment with a CC there's at least a chance of getting some money back... but that's not the only way they extract payment. Even if you made CCs magical anti-fraud devices, there would still be unacceptably high amounts of fraud targeting these types of people, because fraud is a complex problem and the institutions most capable of addressing it systematically would rather live with its predictable costs. As such I don't think adding BTC to the mix would harm things (or help things as a BTC fanatic might try and argue) by a noticeable amount in this area.
Admittedly I'm not too confident on this when applied to fraud more broadly; specifically I'm thinking of the new ransomware threat model people have had to contend with in the last few years. Overwhelmingly those attacks (not just against elders, but industrial infrastructure and so forth!) extract payment with cryptocurrencies, though sometimes they accept gift cards etc. Restricted to the elderly, more traditional computer related schemes like "coerce a phone call to a number for 'technical support' and get victim to grant remote access to their computer, 'discover' malware, have victim pay for 'fixing' it via CC/gift card/bank wire..." are still probably going to be the most common. And of course as I wrote above, by having a payment mechanism that shifts the tradeoff in fraud disputes back to the seller, we might see an increase in seller fraud, but how will that compare to the decrease in buyer fraud?
¹A not so elder relative got tricked out of $120 not too long ago, which for him is a significant chunk of change. Fraud is a multi-spectrum problem and in many respects only seems to be getting worse over time.