If you watch Demystifying Medicine at around the 52 minute mark Dr. Rosenberg says the reason why vaccines have failed is because they don't generate enough T-cells to impact the cancer and you can't get rid of enough Tregs, MDSCs and M2 macrophages . Also, (in my view) he forgot about targets.
To address the first problem GRTS is using a heterologous prime-boost regime . They prime with a single fixed dose of a chimpanzee adenoviral vector and give repeated boosts of a self-replicating, synthetic viral vector. Local Yervoy at a fixed dose is given at the same injection site and an anti-PD-1 (Opdivo in this case) is used as well.
The second can be dealt with via mAbs and small molecules (that deplete and/or reprogramme those immune cells) in development.
As for the third the same company tested its EDGE platform against public methods and looked at how both compared to the tandem minigene approach used by Dr. Rosenberg/NCI . The majority (73%) of MHC class I restricted neoantigens would be included with EDGE vs. only 35% with public tools. They will try to prioritise clonal over subclonal neoantigens  and continue to improve the selection platform with new data. Once the dataset and model for MHC class II restricted neoantigens is sufficiently robust (currently it is ~25%) these will be incorporated into future vaccines as well.
In addition, they have a deal with BLUE .