This paper establishes strong convergence rates for the spatial finite element discretization of a two-dimensional stochastic Navier--Stokes system with transport noise and no-slip boundary conditions on a convex polygonal domain. The main challenge arises from the lack of spatial \(D(A)\)-regularity of the solution (where \(A\) is the Stokes operator), which prevents the application of standard error analysis techniques. Under a small-noise assumption, we prove that the weak solution satisfies \[ u \in L^2\bigl(Ω; C([0,T]; \dot{H}_σ^{\varrho}) \cap L^2(0,T; \dot{H}_σ^{1+\varrho})\bigr) \] for some \(\varrho \in (0,\tfrac{1}{2})\). To address the low regularity in the numerical analysis, we introduce a novel smoothing operator \(J_{h,α} = A_h^α\mathcal{P}_h A^{-α}\) with \(α\in (0,1)\), where \(A_h\) is the discrete Stokes operator and \(\mathcal{P}_h\) the discrete Helmholtz projection. This tool enables a complete error analysis for a MINI-element spatial semidiscretization, yielding the mean-square convergence estimate \[ \|u - u_h\|_{L^2(Ω; C([0,T]; L^2(\mathcal O;\mathbb{R}^2)))} + \|\nabla(u - u_h)\|_{L^2(Ω\times (0,T); L^2(\mathcal{O};\mathbb{R}^{2\times2}))} \leqslant c\, h^{\varrho} \log\big(1 + \frac{1}{h}\big). \] The framework can be extended to broader stochastic fluid models with rough noise and Dirichlet boundary conditions.