... but still needs work.

The IST geometry is plotted below in non-perspective mode.   Figure 1 shows the TGeo model.  Figure 2 shows the STI model exported to TVolume.  Figure 3 shows STI model in perspective.  Some issues are immediately apparent:

1) What is sitting at x=y=0?  It looks to be at about z = +/- 40cm from perspective views.  This is a show stopper for the geometry.
2) The geometry appears to be a reflection  X --> -X.  Is there a bug in going from TGeo/ROOT --> STI?  STI --> TVolume?
3) Looking at the STI in perspective mode... (figure 3)... it looks like there are objects which are "contained" within other objects.  This is a no-no in STI.  The contained volumes will not be seen. 

Figure 1 -- AgML/ROOT model of the IST geometry. 

Figure 2 -- ROOT model exported to STI model, exported to TVolume model

Figure 3 -- Perspective view / wireframe to see masked off volumes...

Below  -- 2.5M points were randomly sampled w/in the IST mother volume.  I plot all points with density > 0.002 (i.e. everthing denser than air).  The points are plotted as R vs Z, with a weight equal to the density of the material found.

What I propose is that we model the IST (using STI) as several cylinders... roughly 13... which would divide the space into regions of similar density.  Below is a 2D histogram which shows the density sampled by throwing points randomly w/in the bounding volume of the IST.

Code which generates it is here:
/star/simu/simu/jwebb/2014/04-04-2014-istreview/GeomAvg.C  (creates TTree)
/star/simu/simu/jwebb/2014/04-04-2014-istreview/DoAvg.C (creates plots)

Location of the output TTree --
/star/simu/simu/jwebb/2014/04-04-2014-istreview/istmate.root

Figure 4 -- R vs Z plot, weighted by density of material in the TGeo Geometry.