Basically, the purpose of the blank is to provide a solution without albumin so that you can measure the absorbance in the absence of albumin to provide a baseline for the rest of the measurements. Ideally, the blank should be the diluent in which the albumin standard is dissolved (for example, distilled water is an appropriate blank if the albumin is dissolved in distilled water).
One problem with this blank, however, is that it will not account for any factors present in the serum which might interfere with the assay. Using the absobance values obtained from the standard to estimate the concentration of albumin in serum requires the assumption that albumin is the only substance present in the serum that will change the absorbance of the solution. In general, this is not a safe assumption to make, but since this method is fairly widely used, I would think that others have tested this assumption and found that it is for the most part true.