This is what I mean. I am referring to sets of particulars, and that there are invariant facts about these sets as long as we form the sets properly. I claim that the sets are manmade, but if the standard of membership is an invariant fact and based on reality, this is fine. Because the standard refers to all known particulars in that set - we cannot refer to things we don't know of - it is fine to call this universal in an epistemic sense.
Now, I don't doubt that if we knew of all particulars, we'd be able to demonstrate a universal in a metaphysical sense. I think this nonsense because there IS no way to know of all particulars. This is not a problem though. It's not subjective either, as it is grounded in the world as it is. Moreover, the set would be manmade anyway, you'd select standards of membership. Again, the standards are invariant facts, so we don't fall into subjectivity. That the definition might change is a personal conflict of how to deal with learning. The facts though? Those don't change.
I don't care if this isn't "really" a universal. I think the word is fine. If not, I don't know a better word.
Overall, I see no issue that what we know to be universal only refers to a delimited set. I don't understand how we -could- refer to all particulars. That would be further from reality.