We consider two issues concerning the dynamics of nanolasers: the impact of their large noise on the deterministic trajectories and the interpretation of the statistical results coming from the photon statistics. The second issue, due to the lack of sufficiently sensitive and fast detectors, represents an unavoidable bottleneck in the analysis of any dynamics and is discussed in detail. Our study reveals several important results. Statistics and dynamics provide complementary information, but the former is not able to reliably distinguish different dynamical regimes. Inferring a coherence emission regime, or degree of coherence, from autocorrelation functions is fraught with interpretative traps. Finally, we address the challenge posed by external noise, which dramatically distorts the statistical information derived from the autocorrelations. We introduce an empirical indicator capable of identifying the lowest pump value from which autocorrelation values become reliable, based on a single sequence of noisy data that can be applied to the same-time, second-order autocorrelation function. This indicator is a valuable tool for experimentalists.