It's not about which API is better, it's about which one is more useful. If Apple were to donate the Metal API to Khronos, I'm sure Metal would become the most useful API because it is the easiest to use.
Why would you think so? I don't believe this to be the case at all. And there are two (and half) main reasons for that.
First, common APIs is that they have to cater to the least common denominator. Metal doesn't. For example, my least favourite aspect of DX12 and Vulkan is how they do resource binding. Metal instead uses structs and pointer manipulation, concepts that are immediately clear to anyone with C or C++ background. The funny thing, AMD's Mantle (the low-overhead API AMD released in 2013 and which in many was served as a precursor to Vulkan) had a nested resource binding model very similar in capabilities to modern Metal. And we know for sure that Intel and AMD support this model because they support Metal. Why doesn't Vulkan or DX12 use this model as well then, but instead implements a more opaque, les flexible one? The official reason stated was that they need to be compatible with different types of hardware, but somehow I doubt that this is because of Qualcomm Adreno or anything like that. Who is the third big GPU vendor that could have put a foot down and blocked these kind of decisions because of their hardware limitation? Exactly.
Now for the second reason. Kronos already had a chance to have a more user-friendly API. They rejected that chance. Vulkan is a new API, and when it was in the design phase Metal, Mantle (and of course the Longs Peaks draft) already existed. But Vulkan instead is an API designed by a small group of driver guys and hardcore low-level gamedev gurus, under tremendous political pressure. Let's not forget that Apple was on the original Vulkan committee. We don't know what happened behind those closed doors, but I am sure that Apple was lobbying to have a more user-friendly API. Then they saw where this is going and decided to pull out.
And for the last part: having an exclusive API is actually good for Apple. They have exclusive hardware which warrants having an exclusive API. This allows them to take the development in the direction they want to take it and not follow the majority. Apple is big enough to do their own things. They can play to the strengths of their hardware and make sure that the devs take advantage of the features properly, rather then hiding them behind extensions that nobody will use.
Of course, this doesn't really help people who want to play games, but this market is not critical for Apple. I doubt that there are many people who are holding off a Mac purchase only because their favourite game does not work on it. And I doubt even more that gamers will start buying Macs in droves once their favourite game works on it.
One thing that Apple can do however is offer a low-level compatibility layer to make porting Vulkan/DX12 apps simpler. I am sure that with just a few low-level tweaks on the Metal side, Vulkan/DX12 style resource binding could become much cheaper to emulate, which would tremendously help projects like MoltenVK and Parallels GPU drivers. And it won't sabotage Metal, since it would not diminish its real capabilities, just improve the compatibility with less flexible "mainstream" binding models.
Edit: to add to this, Apple was successful in bringing Metal roots to the next gen WebGPU API. I would think this is more important, as it influences new generations of developers.