Hubble Frontier Fields (HFF) imaging of the most powerful lensing clusters provides access to the most magnified distant galaxies. The challenge is to construct lens models capable of describing these complex massive, merging clusters so that individual lensed systems can be reliably identified and their intrinsic properties accurately derived. We apply the free-form lensing method (WSLAP+) to A2744, providing a model independent map of the cluster mass, magnification, and geometric distance estimates to multiply-lensed sources. We solve simultaneously for a smooth cluster component on a pixel grid, together with local deflections by the cluster member galaxies. Combining model prediction with photometric redshift measurements, we correct and complete several systems recently claimed, and identify 4 new systems - totalling 65 images of 21 systems spanning a redshift range of 1.4<z<9.8. The reconstructed mass shows small enhancements in the directions where significant amounts of hot plasma can be seen in X-ray. We compare photometric redshifts with geometric redshifts, finding a high level of self-consistency. We find excellent agreement between predicted and observed fluxes - with a best-fit slope of 0.999+-0.013 and an RMS of ~0.25 mag, demonstrating that our magnification correction of the lensed background galaxies is very reliable. Intriguingly, few multiply-lensed galaxies are detected beyond z~7.0, despite the high magnification and the limiting redshift of z~11.5 permitted by the HFF filters. With the additional HFF clusters we can better examine the plausibility of any pronounced high-z deficit, with potentially important implications for the reionization epoch and the nature of dark matter.