Your question is answered in the paper, the protein likely exists as a homodimer in vivo and denaturation (such as performed with SDS PAGE -> Western Blot) separates the dimers into monomers:
"To determine the quaternary structure, size exclusion high-performance liquid chromatography was performed using an Agilent 1100 series high-performance liquid chromatography system with a Bio-Sil SEC-250 column (300 by 7.8 mm) and a mobile phase of 0.1 M Na2PO4, 0.15 M NaCl, and 0.01 M NaN3, pH 6.8. A Bio-Rad standard was used to standardize the column's retention time with respect to molecular mass (supplemental data). All experiments were repeated twice with <0.05 min of deviation in the retention times. The molecular mass of XR was calculated from its retention time to be ∼53 kDa. Monomerization could be induced in the presence of 15% SDS, causing XR to elute as a single peak, with a retention time corresponding to a molecular mass of ∼34 kDa. The data suggest that the native XR is a noncovalently linked dimer. The significant difference between the calculated molecular weight of the dimer and its apparent molecular weight is not uncommon for XRs from other species (7, 21, 33), which also typically exist as dimers. However, to be certain of the protein size, N. crassa XR was subjected to mass analysis by electrospray ionization-quadrupole time of flight mass spectrometry. The highest abundance peak had a value of 38,381 m/z, corresponding exactly with the predicted molecular mass for the His6-tagged XR with the N-terminal fMet removed (supplemental data). Additionally, a 2M+ peak of about 20% abundance at 76,759 m/z corresponds well with the predicted mass of the dimeric form of the enzyme (76,762 Da)."