Which is why they're all so untrustworthy. Take your example: If what actually happened is that you were sitting next to your friend having a beer when a madman came in and shot your friend, and then you wrestled the gun away from the madman and barely escaped with your life, you would get all that same "evidence."
The underlying problem is that we're making decisions that change the course of lives based on weak evidence of any type. Nobody wants to actually follow "beyond a reasonable doubt" because it would mean too low a conviction rate, so instead we suffer too high a false conviction rate.
But you're also confusing what classifies as one type of evidence or the other. The seventeen people who swear you were somewhere else are eyewitnesses to your alibi. In addition, circumstantial evidence is extremely useful to acquit, because although proof that you could have committed the crime doesn't prove that you did, proof that you could not have committed the crime does prove that you didn't. The prosecutor has a tough job, and they're supposed to, because they're proposing to lock a citizen in a cage for a number of years. That's a thing that should happen a lot less often than it does.
You would call me an idiot of I was adamant about the sun rotating around the Earth, but that wasn't always so. Similarly, triangulation has been used for a long time to pinpoint locations, but only now do we know that passive cellular triangulation is actually a singular approximation and so the idea that a deterministic location can be derived from the observation of transmissions is flawed.