Of course there are some errors in real life. The more interesting question is how far from optimality the markets are. To carry on your analogy, a tennis ball is not perfectly spherical but it is round enough for practical use.

Similarly, the market can be 99.99% efficient, in which case it's obviously not entirely perfect but close enough that the average citizen is extremely unlikely to find any real arbitrages (which is what the EMH is essentially about).