As others have commented, kragen is jumping to excitement a bit too quick. GitHub's license data is not curated, and I'm quite sure determining a license of software is an undecidable problem; it just requires human judgment, and many self-report their own license incorrectly to GitHub.
GitHub has a lot of problems to solve before the license data they are presenting can be trusted. Even many very common programs have licenses that can't even be described with an SPDX moniker, so the "badge method" just isn't going to work.
It looks like they try to read `LICENSE`, so maybe if you avoid using the American spelling and just put `LICENCE`, github will skip it. I know gitlab had this bug/feature.
Setting the license manually, of course, would solve the problem.
Also, it seems to be missing a lot of licensing cases. You can have files under multiple licenses in one repository and dual licensed projects. There doesn't seem to be any "GPLv2 only" and "GPLv2 or any later version of the GPL", which is also a big deal. For example Git and Linux are under "GPLv2 only", but are just listed as GPLv2 just the same as they list a license that's "GPLv2 or any later version".
Even more conspicuous is the lack of support for licenses with additional clauses. For example, many compilers like gcc have additional exception clauses allowing you to link the compiler runtime to your library without incurring the GPL requirements. Swift has a similar exception clause for their runtime library. Even though it's Apache v2 which is permissive, Apache v2 still requires distributing the license/copyright info with all copies binary or source for the software, which would be extremely annoying for a compiler to require. LLVM is discussing switching to a similar Apache v2 plus exception as well.
There's also a lot of licenses missing. And I'm not talking about ones that are obscure or equivalent to MIT, I'm talking about for example the Boost license, which is designed specifically to deal with the binary distribution requirements problem, and is popular in the C++ community. [0]
Just looking at the little badge isn't telling you enough useful information. For example, say you are writing a GPLv3 program, and you want to use a library. Looking at the github page shows you its has a GPLv2 badge. But if it's "GPLv2 only" instead of "GPLv2 or any other license" then you can't use it with your project. And the only way you can find that is by reading the license, or at least the README if the author explains the license there. Similarly, just looking at the badge doesn't tell you if the authors have added any additional exception clauses that could be useful in legal compatibility for your program, so you'll have to click through for that as well.
In the end, I think this just provides potential for confusion, and doesn't really make the job of seeing what license the software is under any easier. There could be advantages in listing the license in a standard place and way, but it should allow the developer to manually select the license, and for arbitrary text input for additional notes, or for licenses that github hasn't added. If they insist on keeping it automated, then there should be a way to opt out (of course it'd be best if it was opt-in, but that's not going to happen, so I won't even go there).
0: The copyright notices in the Software and this entire statement, including the above license grant, this restriction and the following disclaimer, must be included in all copies of the Software, in whole or in part, and all derivative works of the Software, unless such copies or derivative works are solely in the form of machine-executable object code generated by a source language processor.