Lagrangian Flow Network (LFN) is a modeling framework in which geographical sub-areas of the ocean are represented as nodes in a network and are interconnected by links representing the transport of propagules (eggs and larvae) by currents. Here we assess the sensitivity and robustness of four connectivity metrics derived from LFN that measure retention and exchange processes, thus providing a systematic characterization of propagule dispersal. The most relevant parameters are tested over large ranges: the density of released particles, the node size (spatial-scales of discretization), the Pelagic Larval Duration (PLD) and the modality of spawning. We find a threshold for the number of particles per node that guarantees reliable values for most of the metrics examined, independently of node size. For the region and dates considered in this study this threshold is about 100 particles per node. We also find that the size of network nodes has a non-trivial influence on the spatial variability of both exchange and retention metrics. Although the spatio-temporal fluctuations of the circulation affects larval transport in a complex and unpredictable manner, our analyses evidence how specific biological parametrization impact the robustness of connectivity diagnostics. Connectivity estimates for long-PLD organisms are more robust against biological uncertainties than for short-PLD ones. Furthermore, our model suggests that for mass-spawners that release propagules over short periods ($simeq$ 2 to 10 days), daily release must be used to properly consider connectivity fluctuations. In contrast, average connectivity estimates for species that spawn repeatedly over longer duration (a few weeks to a few months) remain robust even when using longer periodicity (5 to 10 days). Our results have implications to design connectivity experiments with particle-tracking models and to evaluate the reliability of their results.