Atomistic reconstruction methods are nowadays extremely useful tools to build realistic models of materials and correlates their properties to their atomistic structure. However most of them, like the well-known Reverse Monte Carlo method, consists in building a model able to reproduce some orientationally averaged structural quantities like the Pair Distribution Functions (PDF) of the material, which is usually not discriminatory enough to yield a unique solution for systems having pronounced anisotropic nanotextures. For instance, two drastically different models have been reconstructed from the same PDF of a nanoporous carbon. The first one is a perfectly isotropic pack of small Polycyclic Aromatic Hydrocarbons (PAH) [3] while the second one is a perfect stack of identical infinite graphene sheets containing a few percent of non-hexagonal rings. High Resolution Transmission Electron Microscopy is the technique of choice to describe the nanotexture of aromatic carbons and we have recently introduced a new method, the Image Guided Atomistic Reconstruction (IGAR) method, based on a combination of image analysis and synthesis and molecular dynamics simulations, aiming at building realistic atomistic representations of dense nanotextured carbons. In what follows we briefly recall the methodological details of this approach and how we use it to build models of a rough laminar pyrocarbon, as prepared and after a partial graphitization through heat treatment. The models are then discussed in terms of nanotexture (HRTEM images), local average atomic structure (PDFs) as well as in terms structural defects (sp3 hybridations, non-hexagonal rings. Finally, from these structural observations we end up this paper by proposing some new, high stability, defective graphene sheets.