Using photometric ULTRACAM observations of three new short period cataclysmic variables, we model the primary eclipse lightcurves to extract the orbital separation, masses, and radii of their component stars. We find donor masses of 0.060 +/- 0.008 solar masses, 0.042 +/- 0.001 solar masses, and 0.042 +/- 0.004 solar masses, two being very low-mass sub-stellar donors, and one within 2 sigma of the hydrogen burning limit. All three of the new systems lie close to the modified, optimal model evolutionary sequence of Knigge et al. (2011). We briefly re-evaluate the long-standing discrepancy between observed donor mass and radius data, and theoretical CV evolutionary tracks. By looking at the difference in the observed period at each mass and the period predicted by the Knigge et al. (2011) evolutionary sequence, we qualitatively examine the form of excess angular momentum loss that is missing from the models below the period gap. We show indications that the excess angular momentum loss missing from CV models grows in importance relative to gravitational losses as the period decreases. Detailed CV evolutionary models are necessary to draw more quantitative conclusions in the future.