To be honest, I have not used any of these two software packages since several years ago, but Fluent (now belonging to ANSYS) has been, since many years ago, the leader CFD in aeronautics applications and for computing the aerodynamic performances of wind turbines, racing cars (e.g., Formula-1), etc., because of its superior performance and functionalities. So, if you can choose (because it is very expensive), I would recommend you to use ANSYS-Fluent. Other options are ANSYS-CFX and OpenFOAM (open source, open access, freeware).
Under particular circumstances (like short student projects, undergraduate degree projects, etc.), I recommended to my students using Star CCM+ because at that time it had an automatic mesher, while it was much more time consuming to master Gambit (the mesher with which Fluent used to work).
Fortunately, now there are very good and powerful meshers and automatic meshers that can be used to mesh your geometry. Then you can use these meshes with Fluent, and thus I wouldn't be able to think of any other reason for not choosing Fluent (other than the cost of the license).
I think that a general comparison between two commercial codes is misleading... You should think about the performances of CFD codes in a specified flow problem and using a specific formulation. That would be more meaningful.
In COST Action 732 we compared FLUENT with STAR-CD. I attach a file and you can find in the web page of Action 732 the test cases and the model results. I think that we compared also the codes in COST Action ES1006 and I attach also a file. You can find information in the web page of Action ES 1006.
There is lot of difference in Star CCM+ simulations and Ansys Fluent simulations. But sometimes there is no difference at all! It depends upon how meshing is done and what sorts of boundary conditions are provided while solving.
But still if I want to differentiate between 2 giants in CFD industry, then fluent has more user friendliness while modeling the problem whereas star CCM+ do not provide that much readiness as the media spoke person of CD-Adapco say that ‘user need to know an exact problem so we just provide open environment to the user so he/she can model whichever the way user want’ most of the time Star CCM+ works with Simpson’s 1/3rd rule or 3/8th rule while solving the calculus part of model so the computational cost is more(sometimes you need supercomputer or core cluster computer is must) and solving time is more but results are more accurate without making assumptions. But Ansys uses statistical iterative methods (like Newton-Raphson Iterative method, or In-Situ Adaptive integral Tabulations (ISAT)) which takes less computing cost (you can use laptop or personal computer also) and also less time but results are less accurate and sometimes assumptions.
Star CCM+ uses polyhedral mesh with problem adaptive hybrid cell convergence so that you don’t need to waste time on meshing, but in Ansys meshing is more of kind of open environment so you need to define all sorts of meshing functions like problem and specific cell inflation, cell size etc. otherwise the final solutions may get wrong results.
In the end the user needs to decide what sort of problem he or she needs to model.
I have attached small comparison based article which is not regarding to your research but it may give some directions to your research.
I would suggest you to use Star-CCM+ for pump simulations as the Voith Hydro in Pennsylvania uses Star CCM+ for there Turbomachinery related problems