Design and deterministic spatial arrangement of nanoparticle (NP) clusters are core opportunities and challenges for nanotechnology. Particularly, building functional nanodevices with preset architecture requires to reconcile a high degree of NP organization with a correspondingly accurate placement of the nanocomponents over large areas. Capillary assembly of NPs (CAN) combines the complementary advantages of bottom-up chemical synthesis of NPs with those of top-down nanofabrication, ideal for precise definition of topographic features arbitrarily distributed over large surfaces. However, in spite of remarkable previous insights, complete understanding of the technique is still elusive. In this work, we identify three sequential stages in the dynamics of CAN - namely, insertion, resilience and drying - and demonstrate that each stage distinctively affects the cumulative yield of the process. We then propose an effective trap design, tailored in all three dimensions, to achieve an optimal assembly yield, and highlight its performance through the assembly and characterization of plasmonic nanoantennas.