In general, you will want to use a distance large enough to ensure every feature has at least one neighbor. An asterisk indicates a distance that does not ensure all features will have at least one neighbor. Some tools such as Cluster and Outlier Analysis and Spatial Autocorrelation require each feature to have at least one neighbor.
If a smaller distance is more appropriate (if it better matches the scale of the question you are asking), you can create an SWM file using the Generate_Spatial_Weights_Matrix tool and set the Number of Neighbors parameter to the minimum number of neighbors each feature must have. If, for example, you set the minimum to two, the search distance for any feature with less than two neighbors will be expanded, only for that feature, in order to capture its two nearest neighbors.