Complex pathways in spontaneous molecular assemblies have been envisaged to be crucial factors in determining diverse outcomes desirable for materials design. Computer simulations can be instrumental in elucidating early stages of the assembly process, where direct experimental observations are difficult to obtain. In this paper, we report a computational study on the pathway complexities in aqueous mixtures of hydrophobic tripeptides that self-assemble into amorphous nanoscale objects. We introduce a new algorithmic approach for "complete accounting" of the simulation data in terms of rearrangements of spatial clusters spanning multiple length and time scales without relying on a priori assumptions. The salient features of an assembly mechanism were analyzed by focusing on four types of "events" encompassing all possible redistribution of molecules among interacting clusters. The event tracking analysis of the dynamics of spontaneous assemblies observed in fully atomistic molecular dynamics trajectories of tri-phenylalanine, tri-leucine, and tri-isoleucine revealed non-classical pathways with a clear trend in terms of spatio-temporal hierarchy. The discernible differences between latter two systems, where the building blocks are constitutional isomers, were particularly striking. Tri-phenylalanine assembled by a strongly hierarchical mechanism involving fusions of clusters of varying sizes. On the other hand, tri-isoleucine mostly followed a ripening type growth process, where isolated molecules or small molecular clusters attached to a single growing cluster. The tri-leucine exhibited an intermediate behavior with subtle differences from other two systems. These results underscored the effectiveness of our new analysis protocol in deciphering pathway complexities in molecular assemblies hitherto unreported in the literature.