Can PIL tests be performed with the same code generated with normal build process?

I use Simulink Embedded Coder to generate C code from models to be integrated as separate modules in the SW build of our targets. FreeRTOS and VxWorks are the OS of our target devices. I use PIL tests to verify if math libraries and the algorithm in general behaves in the same way on the target processor, using the same unit tests run on the PC implemented with the MATLAB unit test framework. Because of our SW architecture, the C code generated from Simulink models for integration into the target is different from that one automatically generated for PIL testing. The efforts spent in setting up PIL tests is remarkable and in my current view it is a shame not being able to test the same code integrated in the target. Is there a way to force the PIL automatically generated code to be the same code generated with the standard build process? I know that VectorCAST proposes some solutions to this problem, where the input and expected output datasets can be imported from external files. Should I instead consider the generated C code an intermediate product that I should not consider in terms of consistency of the tests? Thanks.