This might or might not be a bug, but despite your cool "try to interprete all inputs" agenda I do recommend to reject both(!) terms. Reason being that this sort of stuff is one of three things: 1st: Someone testing around. Well, look what we got. 2nd: Someone being stupid. Well, maybe you might guess the intended input or maybe not, but really at this point it's stupid enough so you should tell the user to not do this this way.
3rd: Someone inputs something big and serious and has one or multiple typos. This is the most important case and not interpreting this might save this guy! If this is in some big equation and the typo was missed once, then it is likely to be missed again in the input representation field which will often not perfectly match the actual input anyway.
If you absolutely disagree with me, then you might still look at the input representation of https://www.wolframalpha.com/input/?i=tan%5E-%281*2%29+%28x%29. It looks wrong.
Edit: Despite reading FAQ and rules I don't know if I posted this on the right place. If wrong, then I hope that you can move it to the correct place. FAQ should tell this!