This paper proposes two contributions. One is an analysis for the limit of the subject of goal-directed reactive robot navigation, and the other is an effective navigation method employing the scheme of the subject. The analysis for the subject is presented in order to clarify the limit of the method. On the basis of the analysis, a safety-guaranteeing and deadlock-free reactive navigation method is newly proposed. The proposed method has a simple behavior-based frame such that it can make the required navigation tasks such as obstacle avoidance, deadlock resolving, and etc. with a very small set of behaviors in entirely unknown environments such as a living room, an office, and etc. Some results of experiments show these validities.