Anastylosis is an archaeological technique which focuses on the reconstruction of collapsed building and destroyed artworks, starting from the original pieces. Many digital approaches have been developed in the last decade, mainly based on 2D and 3D analysis of the structure of the fragments. These techniques aim at supporting the priceless work of the involved operators, mainly in the decision processes and in the resolution of positioning ambiguities. Techniques acting with this scope lie in the field of the digital anastylosis. In this paper we present SAFFO, a digital approach to 2D reconstruction of frescoes, based on the extraction of SIFT features from a painting. The approach appears to be very robust to false positives, resulting optimal in scenarios involving fragment sets containing spurious elements. The experiments have been performed on the DAFNE (Digital Anastylosis for Fresco challeNgE) dataset, which gathers more than 30 2D artworks and provides several tessellation for each. For its robustness against spurious fragments, SAFFO won the third place in the rank list of DAFNE Challenge 2019.