We perform a comprehensive study of the total mass distribution of the galaxy cluster RXCJ2248 ($z=0.348$) with a set of high-precision strong lensing models, which take advantage of extensive spectroscopic information on many multiply lensed systems. In the effort to understand and quantify inherent systematics in parametric strong lensing modelling, we explore a collection of 22 models where we use different samples of multiple image families, parametrizations of the mass distribution and cosmological parameters. As input information for the strong lensing models, we use the CLASH HST imaging data and spectroscopic follow-up observations, carried out with the VIMOS and MUSE spectrographs, to identify bona-fide multiple images. A total of 16 background sources, over the redshift range $1.0-6.1$, are multiply lensed into 47 images, 24 of which are spectroscopically confirmed and belong to 10 individual sources. The cluster total mass distribution and underlying cosmology in the models are optimized by matching the observed positions of the multiple images on the lens plane. We show that with a careful selection of a sample of spectroscopically confirmed multiple images, the best-fit model reproduces their observed positions with a rms of $0.3$ in a fixed flat $Lambda$CDM cosmology, whereas the lack of spectroscopic information lead to biases in the values of the model parameters. Allowing cosmological parameters to vary together with the cluster parameters, we find (at $68%$ confidence level) $Omega_m=0.25^{+0.13}_{-0.16}$ and $w=-1.07^{+0.16}_{-0.42}$ for a flat $Lambda$CDM model, and $Omega_m=0.31^{+0.12}_{-0.13}$ and $Omega_Lambda=0.38^{+0.38}_{-0.27}$ for a universe with $w=-1$ and free curvature. Using toy models mimicking the overall configuration of RXCJ2248, we estimate the impact of the line of sight mass structure on the positional rms to be $0.3pm 0.1$.(ABRIDGED)