Rather than doing a variable-sample rate FFT one could simply export the relevant data at the desired sample rate. This is compatible with a variable-step solver - it only ensures that simulation solutions are reported at the specified time step. This would only be problematic if the sample time was set excessively small.
One could use the PLECS "To File" block or one of the many logging options included in Simulink (the "To Workspace" block comes to mind). Both of these blocks have a "Sample time" parameter.
If you want to proceed using variable-sample rate data, could you share your script or a simple example? The nufft() results are sensitive to the input arguments.