> The program runs without error, however there is no eject during the merge. 

Typical ejecta from BNS mergers amount to a very tiny fraction of the total initial mass, with values of 1e-3 Msun from BNS merger simulations being reported in the literature (https://arxiv.org/abs/1809.11161). This value is highly dependent on the equation of state however. Further most simulations reporting ejecta are not performed at multiple numerical resolutions, meaning that the values may be adjusted downwards with subsequent simulations. Indeed in most simulations I saw for which more than one resolution was performed in that paper, the higher resolution simulation has less ejecta (see Table 2)--sometimes significantly less (e.g., LS220_M140140_LK).

Bottom line, I am not surprised you didn't see any ejecta. Ejecta measurements are often dominated by numerical error, and depend sensitively on equation of state. The very simple equation of state you chose (simple Gamma=2 polytrope) might not exhibit much, if any, ejecta in the limit of very high resolution. 

A recent update to IllinoisGRMHD supports more sophisticated (piecewise polytrope/"hybrid") equations of state. It exists within the IllinoisGRMHD subdirectory of https://github.com/zachetienne/nrpytutorial . You might have more luck getting ejecta from BNS initial data with piecewise-polytrope equations of state.

> And "restmass. Init. IN sphere @ (0.000000e+00,0.000000e+00,0.000000e+00), r=2.708902e+02. Moves/Tracks AMR Centre -1/-1" is increasing.

If I'm interpreting this correctly (I might not be), you are measuring total rest mass within 270.89 in code units. You seem to observe an increase of 0.06% over the course of the simulation. That's not much, and can happen due to interpolation errors at AMR refinement boundaries (matter crosses AMR refinement boundaries and tiny errors add up to a small boost in mass), or become completely unreliable at the time of black hole formation. Performing another simulation at higher or lower resolution and additional volume integral regions may help diagnose this measurement. I think you'd want to analyze the rest mass *outside* an interior volume to estimate ejecta anyway. 

A more sophisticated interpolation treatment at AMR refinement boundaries might help reduce this error, which amounts to a different prolongation type (e.g., ENO) being chosen for evolved GRMHD variables.

Hope this helps.