Traffic microsimulation is central to smart-city transport control because digital twins and decision-support platforms depend on well-calibrated behavioural models before operational strategies are tested in the field. This paper examines how objective function design shapes calibration efficiency and parameter convergence consistency in a detector-based microsimulation setting. The study uses a VISSIM model of a 2.5 km weaving segment on the southbound Antwerp R1 motorway and calibrates 41 driving-behaviour parameters using detector-level speed and headway observations from the morning peak of 10 September 2019. A multifaceted objective function based on the 1-Wasserstein distance is evaluated against the Kolmogorov–Smirnov (K–S) distance and root mean squared relative error (RMSRE), under single-KPI and dual-KPI formulations. Two optimisers are considered: simultaneous perturbation stochastic approximation (SPSA) as the primary high-dimensional method and Bayesian optimisation as an external validation benchmark. In the synthetic experiment, the random-seed noise effect is 1.2%, and the optimisation trajectories approach attainable minima near 1.75% (SPSA) and 1.6% (Bayesian optimisation) when starting from dispersed initial points. In the real-data experiment, the strongest and most balanced behaviour is obtained when the Wasserstein distance is paired with a speed-plus-headway objective. Across the 32 dominant-class parameter instances used for convergence assessment, this setting yields 13 parameters meeting the consistency threshold under SPSA, compared with 5 under the K–S speed-plus-headway formulation. Although speed-only RMSRE also stabilises 13 parameters, it does so under a single-KPI setting that does not preserve balanced performance across KPIs. The results show that, for smart-city traffic digital twins, calibration quality depends not only on the optimiser but on whether the objective function preserves traffic-state heterogeneity and constrains the parameter search with sufficiently rich behavioural information.