We investigate the relationship between the halo mass, M_200, and concentration, c, for a sample of 26 group- and cluster-scale strong gravitational lenses. In contrast with previous results, we find that these systems are only ~ 0.1 dex more over-concentrated than similar-mass halos from dark matter simulations; the concentration of a halo with M_200 = 10^14 M_sun is log c = 0.78±0.05, while simulations of halos with this mass at similar redshifts (z ~ 0.4) predict log c ~ 0.56 - 0.71. We also find that we are unable to make informative inference on the slope of the M_200-c relation in spite of our large sample size; we note that the steep slopes found in previous studies tend to follow the slope in the covariance between M_200 and c, indicating that these results may be measuring the scatter in the data rather than the intrinsic signal. Furthermore, we conclude that our inability to constrain the M_200-c slope is due to a limited range of halo masses, as determined by explicitly modelling our halo mass distribution, and we suggest that other studies may be producing biased results by using an incorrect distribution for their halo masses. ; Comment: 8 pages; accepted to MNRAS