I agree. And it's pretty obvious the reason why is that Freenet presents itself as a web server. Any privacy project packaged as a web app is doomed. The local web app is a shitty interface. If Freenet had an interface more like Bittorrent, its fate may have been different.
There is a market for appliances sporting web interfaces, but that's only because it's the cheapest way to have a graphical interface that's cross platform. Only technical people use such things, and even they have to want the benefits of the appliance pretty badly to put up with the hassle of deployment. The average consumer will never touch that crap, and I don't blame them.
You mention consumer routers. Yes, those are appliances with web interfaces. But the benefit of deploying one is ACCESS TO THE INTERNET! That's a pretty compelling reason. Even if LookingGlass was sold pre-configured like a router, it's not even close to compelling enough to gain traction.
Besides how inconvenient the web appliance is to use, it's also pretty weak from a security perspective. You want to have all the encryption happening as close to the endpoints as possible. To reduce attack surface it's also best that the architecture is as simple as possible, with minimal dependencies. The LookingGlass approach follows neither of those best practices. The project even recommends deployment behind a firewall.
I don't want to shit on LookingGlass here. I haven't looked too hard at it and I support any experimentation with new protocol ideas in this space.
It's just that to be successful it needs to be natively written for every platform, especially the mobile ones. Making the core implementation portable by using C or C++ is acceptable. Just getting it going on Linux and asking users to run a server - especially a physical server - is totally not.