If they have the allocation, why wouldn't they use it?
That's how IP was supposed to work. And once you've worked in a network where everyone has a real, non-RFC1918 address, you'll probably not want to go back. It just makes things really convenient; you never need to worry about address space collisions if you decide to connect two building networks together, for instance ... it's a whole lot more elegant than private addressing.
The places I've worked at that have legacy allocations have absolutely no interest in switching to RFC1918, and their allocation agreements are old enough that they don't contain any clawback provisions. It'll be "from my cold, dead hands" before they transition, and I can't blame them. It's how the Internet was supposed to work.
It's unfortunate that IPv6 got used as a way to shoehorn in a lot of stuff besides just increasing the address space. I've periodically seen much more conservative proposals that would have been less of a "forklift upgrade" and I think in retrospect it's unfortunate that these weren't considered.
But if IPv6 manages to kill or at least discourage the widespread use of RFC1918 addresses and bring back a more peer-to-peer architecture, it will be worth it in the long haul. The variety of applications and use cases we would be killing off without even knowing it by continuing down the path towards multiple layers of private addresses (because they just wouldn't be possible under that topology, and hence would never get developed) ought to sober anyone interested in development and new technology.
It may be possible for the Internet of today to work just fine with ISP-level NAT on top of home-network NAT, but if we resign ourselves to that we're shutting out a whole lot of interesting future possibilities that can only happen with an Internet that has end-to-end routability without second-class citizens. That would be truly sad.