As far as I can tell, this paper is based on a simulated statistical epidemiology model, not an observed clinical trial. There is a difference, especially when you are trying to model THIRTY YEARS in the future. That is eons of time in terms of biostatistics. For medical instance events that occur with such fine granularity as shingles and chicken pox, predicting what will happen at (essentially) asymptotic boundaries is difficult.
There are no people who received the chicken pox vaccine 31 years ago. Zero cases exist for study. This paper consists of what is essentially a good guess about medical outcomes 20, 30 years in the future. Predictive epidemiology models never hold as much credence as observed studies.