Xylanases Tx-xyn10 and Tx-xyn11 were compared for their transxylosylation abilities in the presence of various acceptors. Tx-xyn10 exhibited a broad specificity for various acceptors, whereas xylanase Tx-xyn11 catalysed transxylosylation reactions only in presence of polyphenolic acceptors. A modelling approach was developed to study the molecular bottlenecks into the active site of the enzyme that could be responsible for this restricted specificity. The glycosyl-enzyme intermediate of Tx-xyn11 was modelled, and a rotamer of the Y78 residue was integrated. In silico mutations of some residues from the (+1) and (+2) subsites were tested for the deglycosylation step in the presence of non-polyphenolic acceptors. The results indicated that the mutant W126A was able to use aliphatic alcohols and benzyl alcohol as acceptors for transxylosylation. Experimental validation was tested by mutating the xylanase Tx-xyn11 at position W126 into alanine. The specific activity and catalytic efficiency of the W126A mutant during the hydrolysis of xylans decreased by 2-fold and 4-fold, respectively, compared to wild-type xylanase. Among tested acceptors, transxylosylation catalysed by mutant W126A was improved with benzyl alcohol leading to a 2-fold higher concentration of benzyl xylobioside, as predicted by in silico mutation. This improved transxylosylation in the presence of benzyl alcohol leading to higher synthesis of benzyl xylobioside could likely be explained by lowest steric hindrance in the aglycone subsite of the mutated xylanase. No secondary hydrolysis of benzyl xylobioside occurred for both wild-type and mutant xylanases. Finally, our results demonstrated that the modelling approach was limited and that accounting for protein dynamics can lead to improved models.