We compare dynamical mass estimates based on spatially extended stellar and ionized gas kinematics (M_dyn,*_ and M_dyn,eml_, respectively) of 157 star-forming galaxies at 0.6<=z<1. Compared with z~0, these galaxies have enhanced star formation rates, with stellar feedback likely affecting the dynamics of the gas. We use LEGA-C DR3, the highest-redshift data set that provides sufficiently deep measurements of a Ks-band limited sample. For M_dyn,*_, we use Jeans anisotropic multi-Gaussian expansion models. For M_dyn,eml_, we first fit a custom model of a rotating exponential disk with uniform dispersion, whose light is projected through a slit and corrected for beam smearing. We then apply an asymmetric drift correction based on assumptions common in the literature to the fitted kinematic components to obtain the circular velocity, assuming hydrostatic equilibrium. Within the half-light radius, Mdyn,eml is on average lower than M_dyn,*_, with a mean offset of -0.15+/-0.016dex and galaxy-to-galaxy scatter of 0.19dex, reflecting the combined random uncertainty. While data of higher spatial resolution are needed to understand this small offset, it supports the assumption that the galaxy-wide ionized gas kinematics do not predominantly originate from disruptive events such as star formation-driven outflows. However, a similar agreement can be obtained without modeling from the integrated emission line dispersions for axis ratios q<0.8. This suggests that our current understanding of gas kinematics is not sufficient to efficiently apply asymmetric drift corrections to improve dynamical mass estimates compared with observations lacking the signal-to-noise ratio required for spatially extended dynamics.