Top: count of retained NS events within an event sample constructed using a given p_lim, rescaled to yields from the full GBTDS footprint. The error bars representing Poisson noise are the same as in Fig. 2. Centre: recall of NS events, calculated as a ratio of selected NS events to all detectable NS events. Bottom: purity of the sample, calculated as a ratio of selected NS events to all events selected for the sample. For higher p_lim, values for v_kick ∈ (150, 250) km/s are missing as samples cannot be constructed (i.e. the resulting sample size is 0).