The synthesis of disaccharides, particularly those containing hexofuranoside rings, requires a large number of steps by classical chemical means. The use of glycosidases can be an alternative to limit the number of steps, as they catalyze the formation of controlled glycosidic bonds starting from simple and easy to access building blocks; the main drawbacks are the yields, due to the balance between the hydrolysis and transglycosylation of these enzymes, and the enzyme-dependent regioselectivity. To improve the yield of the synthesis of β-d-galactofuranosyl-(1→X)-d-mannopyranosides catalyzed by an arabinofuranosidase, in this study we developed a strategy to mutate, then screen the catalyst, followed by a tailored molecular modeling methodology to rationalize the effects of the identified mutations. Two mutants with a 2.3 to 3.8-fold increase in transglycosylation yield were obtained, and in addition their accumulated regioisomer kinetic profiles were very different from the wild-type enzyme. Those differences were studied in silico by docking and molecular dynamics, and the methodology revealed a good predictive quality in regards with the regioisomer profiles, which is in good agreement with the experimental transglycosylation kinetics. So, by engineering CtAraf51, new biocatalysts were enabled to obtain the attractive central motif from the Leishmania lipophosphoglycan core with a higher yield and regioselectivity.