Atmospheric Phase Screens (APSs) derived from Interferometric Synthetic Aperture Radar (InSAR) observations contain the difference between the tropospheric water vapor induced delay of two acquisition epochs, i.e. the slave and the master (or reference) epochs. Using estimates of the atmospheric state coming from independent sources, for example numerical models and/or Global Navigation Satellite System (GNSS) observations, the APSs can be transformed into absolute maps of Tropospheric Delay (Zenith Total Delay or ZTD), related to the columnar atmospheric water vapor content. In this work, a systematic comparison between various APS and ZTD products aims to determine a convenient strategy to go from APSs to InSAR-derived absolute ZTD maps, highlighting the uncertainties and approximations introduced in the entire processing. The main problem to solve is the evaluation of a sufficiently accurate high resolution master delay map. Different sources of data and two different approaches to derive the master are validated and compared to define the most suitable strategy for meteorological applications. Maps of ZTD obtained by an iterative interpolation of a global atmospheric circulation model values result more suited than those derived from the assimilation of GNSS observations into a NWP model. A time average approach to estimate the master map is more robust than the single epoch approach with respect to the choice of the master epoch. Still the choice of a proper master epoch in the InSAR processing chain as well as that of the maps to be averaged result crucial in the estimate of the master.