Embedded coder produces incorrect code reading from Simulink data store

I have a shared structure that holds the coefficients of a biQuad filter block. This is for a 10th order filter that is implemented as 5-2nd order sections (SOS). The structure is defined as:

typedef struct {
  real_T      numerator[3][5];
  real_T      denominator[2][5];
  real_T      gain[6];
} userFilterSOS5;

typedef struct 
{
  int16_T      BP_LPsel;  // [0 = DC, 1 = BPF] connects to LP input
  int16_T      LP_HPsel;  // [0 = DC, 1 = BPF, 2 = LPF] connects to HP input
  int16_T      HP_notchSel;  // [0 = BPF, 1 = LPF, 2 = HPF] connects to notch input
  int16_T      output_Sel;  // [0 = BPF, 1 = LPF, 2 = HPF, 3 = notch] connects to output
  int32_T      DCdelay;
  int32_T      BPdelay;
  int32_T      LPdelay;
  int32_T      HPdelay;
  int32_T      N1delay;
  int32_T      N2delay;
  int32_T      N3delay;
  int32_T      N4delay;
  real_T      FIRbpfCoef[301];  // FIR filt has only numerator coefs
  userFilter7    LPF;
  userFilter7    HPF;
  userFilterSOS5  notch1;
  userFilterSOS5  notch2;
  userFilterSOS5  notch3;
  userFilterSOS5  notch4;
} filtersBus;

It is declared as:

filtersBusSig = Simulink.Signal;
filtersBusSig.DataType = 'Bus: filtersBus';
filtersBusSig.SamplingMode = 'Sample based';
filtersBusSig.Complexity = 'real';
filtersBusSig.InitialValue = 'filtersBusStruct';
filtersBusSig.StorageClass = 'ExportedGlobal';

In the Simulink DataStoreRead block, it is specified as:

filtersBusSig.notch1.numerator(:,:)
filtersBusSig.notch1.denominator(:,:)
filtersBusSig.notch1.gain(:,:)

The embedded decoder generates the following code to read the data out of the DataStore:

      /* DataStoreRead: '<S22>/notch1Coef' */
      for (i = 0; i < 15; i++) {
        decoder_decBPFpump_B.notch1Coef_o1[i] = filtersBusSig.notch1.numerator[i];
      }

We have tried this with the Visual Studio 2013 compiler, Intel compiler, and QtCreator compiler, and all of them generate something like the following error:

../decoder-common-lib/decoder_Main_Model/testRepo/slprj/ert/decoder_decBPFpump/decoder_decBPFpump.c:
756: error: incompatible types when assigning to type ‘real_T’ from type ‘real_T *’

I understand that the embedded coder likes to convert 2D arrays to 1D vectors, but this code will not compile. Is there a way to correct this, or at least force it to generate a warning?

I should add that converting the arrays to 1D vectors is not a solution, since the block generates an error if the inputs are not in (3, N) and (2, N) format for the numerator and denominator respectively.