We study the orientation dependence of the reconstructed deceleration parameter as a function of redshift. We use the Union 2 and Loss datasets, and the well known preferred axis discussed in the literature, and find the best fit reconstructed deceleration parameter. Our results show that a low redshift transition of the reconstructed q(z) is clearly absent in one direction and amazingly sharp in the opposite one. We discuss the possibility that such behavior can be associated to large scale structures affecting the data.