Short answer: Literally speaking we can't have perfect knowledge. Practically speaking we can.
Long answer: All knowledge should be amendable to future evidence, but that doesn't mean there aren't things that we can say we know for sure - mathematics for one. There are things that have been established to such a high degree it would take such massive evidence to change them that it's a pretty safe bet they won't change any time soon and it's an even safer bet that if they do, it won't involve completely throwing out the old theory - Newton's laws are horribly outdated, but they still work well enough so we don't throw them out.
Even if Hydrogen was two functionally identical atoms that couldn't be discerned using any method we know, the periodic table makes detailed, useful predictions, all of which are fulfilled 100% of the time. That means it's either the truth and we know about how elements behave, or it's a perfectly predictive model of how elements behave, indistinguishable from the truth. At that point, any difference is so insignificant, we can't distinguish between the two so for all intents and purposes, we can just say the model and the truth are one in the same. Hence for practical purposes, we can know things.
To put it mathematically, what's the difference between 1 and 0.999...recurring?
The answer: none, they're mathematically identical. Similar principle.