Hi fikkser,
In your example, I'm assuming you used center point focusing, and the camera focused on the highest contrast border within the AF sensor area, the back edge of the lens cap. It will also bias towards a contrast border that is closer to perpendicular to the direction to either of the cross lines of the sensor. You are correct in assuming that the center focus area is approximately the size of the central ( ) in the VF.
From your example, I'd call this user error. Eliminate the lens cap from the sensor area by actuating AF with the camera pointed a touch higher, lock focus on the bottle, recompose, then shoot. As the photographer, it's your responsibility to make sure the camera is doing what you want it to, not leaving it to chance.
This one took multiple tries at AF before I could take the shot.
This is not to say that your lens/camera combo might not be correctly adjusted to each other -- but if the gear works properly with flat surfaces as you mentioned, then it points to a problem in your technique. . .
Scott