Secure boot exists primarily because it provides a "feel good" answer to content providers about the feasibility of DRM. It's not an answer to real world security issues and never has been.
If you want the OS to trust the hardware it's on, you need a TPM, and you probably need the machine to be installed by someone who can communicate with the content provider. The entire point of Secure Boot is that it can be implemented just in existing UEFI code, _without_ the additional hardware of a TPM.
I develop an OS product that has no browser installed on it at all -- there's one small client application that just interacts with our one server product, and then boots everything else you might want to do in a VM. The biggest threat to the security of our outer layer is, in fact, an evil maid attack. It's certainly worth being aware of that threat, even if there are other, bigger threats. (Nobody says that we should stop worrying about buffer overflows while there are still cross-site scripting attacks, etc.)
That's a powerful sales tool. Systems that don't have secure boot have a much harder time selling their DRM implementations. And that's why we have secure boot. Don't fool yourself into thinking all problems are technical.
Given that this has not been done, and given the small number of people this scenario interests, given that thoses people are unlikely to be interested by Windows 8 to begin with, and even less by Windows RT, Secure Boot has not been designed by MS to prevent that, but really to prevent the installation of other operating systems. The FSF just fight Secure Boot for the real reason it has been created; if you create and alternate security feature for the "evil maid" scenario that preserve user freedom, that would be a completely different subject and you want angry the FSF with it. The FSF's attitude towards new technologies is not reactionary. They fight for user's freedom, as always.
The only way to defeat evil-maid attacks is for the early-boot code in your computer's firmware to cryptographically validate the bootloader or kernel that it's about to start, and make sure it's signed by a trusted authority, and make sure that it's possible for the computer owner to disable the ability for anyone other than him to change the set of trusted authorities. This happens to be exactly what Secure Boot does.
Secure Boot where you can only run Microsoft-approved software is bad for software freedom, no question about it. Secure Boot as defined by the spec, where there is no requirement to have preloaded MS keys and where users explicitly can load their own keys, is absolutely _good_ for software freedom -- it makes the "evil maid" attack much, much harder. (The "evil maid" attack on full-disk encryption involves someone having access to your computer while it's off, and trojanning the bootloader and having it leak your disk password or the contents of the disk once your system boots. People naively expect full-disk encryption to stop this, but it's as easy as booting from a CD or thumbdrive. With secure boot, it involves physically opening the case and probably replacing chips.)
Let's make sure that Secure Boot stays good for software freedom, instead of throwing out the technology along with the policy, and leaving users with no defense against this sort of attack.