Casper has hit the nail on the head here with regards to using TransDem for European and UK routes, there isn't a great deal of high quality Dems available for the area.
Unless you are in the Nullarbor there is nothing like flat terrain in the real world. Plane sea-level baseboards are in invention by Trainz and have no prototypical counterpart. We might not have 20m or better DEMs for free outside North America, but we do have SRTM and GDEM. Jerker has said it. In any case some DEM will always make for a more realistic route than no DEM.
There is nothing wrong with a "Basemap" approach, though, as has been mentioned before. It is one one aspect and one of the established techniques of prototypical route building. There are other aspects and there are other approaches. One might even combine some of the established techniques.
It’s probably too much to expect to get a perfectly finished result from DEM software. However, could I expect to get as good a result, or better, with TransDem utilising the Japanese data, than the unfinished terracing results shown in this shot?
Why do you think DEM software cannot achieve better results? If the original DEM does not have terraces, TransDEM created terrain will not have either. Much has been said elsewhere (not Trainz related) about artefacts in ASTER data. ASTER was available before SRTM and the artefacts too. That's why I implemented some specific algorithms to deal with them, quite early in TransDEM development.
I would always compare SRTM with ASTER. SRTM 3 arc sec is not so bad at all. Better than the terrain in your screenshot. There are other options, too. For the European Alps we have ViewfinderPanoramas 1 arc sec and for the UK we have O/S Land-Form Panorama 50 m.