Nanobodies have been progressively replacing traditional antibodies in various immunological methods. However, the use of nanobodies as capture antibodies is greatly hampered by their poor performance after passive adsorption to polystyrene microplates, and this restricts the full use of double nanobodies in sandwich enzyme-linked immunosorbent assays (ELISAs). Herein, using the human soluble epoxide hydrolase (sEH) as a model analyte, we found that both the immobilization format and the blocking agent have a significant influence on the performance of capture nanobodies immobilized on polystyrene and the subsequent development of double-nanobody sandwich ELISAs. We first conducted epitope mapping for pairing nanobodies and then prepared a horseradish-peroxidase-labeled nanobody using a mild conjugation procedure as a detection antibody throughout the work. The resulting sandwich ELISA using a capture nanobody (A9, 1.25 μg/mL) after passive adsorption and bovine serum albumin (BSA) as a blocking agent generated a moderate sensitivity of 0.0164 OD·mL/ng and a limit of detection (LOD) of 0.74 ng/mL. However, the introduction of streptavidin as a linker to the capture nanobody at the same working concentration demonstrated a dramatic 16-fold increase in sensitivity (0.262 OD·mL/ng) and a 25-fold decrease in the LOD for sEH (0.03 ng/mL). The streptavidin-bridged double-nanobody ELISA was then successfully applied to tests for recovery, cross-reactivity, and real samples. Meanwhile, we accidentally found that blocking with skim milk could severely damage the performance of the capture nanobody by an order of magnitude compared with BSA. This work provides guidelines to retain the high effectiveness of the capture nanobody and thus to further develop the double-nanobody ELISA for various analytes.