The realisation of hyperspectral terahertz imaging is a significant step towards understanding of the life sciences on all scales. A key to this understanding is the retrieval of dielectric properties from such images, a task which is plagued by experimental limitations, challenging the terahertz community for more than two decades. In this contribution, we propose a new combined retrieval methodology to overcome misalignments and Fabry-Pérot effects on the extraction of the dielectric properties of human bone samples through the combination of the Kramers-Kronig relations and Fabry-Pérot reflection modelling. Results extracted from ∼100 µm human bone slices composed largely of collagen are consistent with those measured for pristine collagen samples. This represents another stepping-stone towards the adoption of terahertz imaging into pre- and clinical practice.