How much can we trust the electromechanical delay estimated by using electromyography?