Fabel, you are confusing the fallibility of humans with whether the game itself is solvable and therefore capable of being played "perfectly". A computer will eventually be able to play the perfect game. We aren't there yet, but in about 200-225 years, we will be.
Tic-tac-toe is solvable. It's been solved. But a human could get careless and make a mistake if they had to play 100 games of tic-tac-toe simultaneously. That doesn't change the nature of the game.
"If you point out that someone got "lucky" in a single game, that merely means his opponent didn't find the best moves"
Portisch won an unwinnable endgame because his opponent was mistaken about the time control. Such things happen, but one can never say that Reshevsky could have beaten Geller in the same way the following round by showing the same skill. Some results will always be influenced by random factors, regardless if chess is solvable, as long as humans are involved in the game.