By applying compressive sensing, we can obtain radar range profile using only part of the frequency bins in step-frequency radars. While the probability of success in obtaining the range profile depends on which frequency bins are used, it seems that there is no general and simple method to choose the set of frequency bins for better probability of success. Although random selection is said to be good enough, it would be better if we select those in some strategic ways. This letter deals with frequency selection strategies when deterministic ways are not applicable and the super-resolution range profile is necessary. The proposed strategies are tested with simulations and real measurement data set. The results show that the proposed method is better in achieving the probability of exact recovery than casual random selections.