Electronic – Problem with SPI communication (PIC24H)

I'm trying to communicate with a NRF24L01 transceiver module with a PIC24HJ128GP202. Besides all the code, here is the important part:

void NRF_Config()
{
  NRF_WriteRegister(0x02, 0x01); //PIPE 0 ENABLE
  NRF_ActivateSR();
  NRF_WriteRegister(0x1D, 0x06); //FEATURE REG - Enable Dynamic Payload && Enable ACK Payload
  NRF_WriteRegister(0x1C, 0x01); //ENABLE DYNAMIC PAYLOAD PIPE0
  NRF_WriteRegister(0x00, 0x5A); //CONFIG REG
  Delay(Delay_2ms);
  NRF_EN(); //CHIP ENABLED PIN = TRUE
}

with the WriteRegister function being:

unsigned char NRF_WriteRegister(unsigned char reg, unsigned char data)
{
   SPI1_BEGIN(); //CHIP SELECT = LOW  (START COMMUNICATION)

   IFS0bits.SPI1IF = 0;
   while (SPI1STATbits.SPITBF);
   WriteSPI1(0x20 | reg);
   while (!IFS0bits.SPI1IF);

   IFS0bits.SPI1IF = 0;
   while (SPI1STATbits.SPITBF);
   WriteSPI1(data);
   while (!IFS0bits.SPI1IF);

   SPI1_END(); //CHIP SELECT = HIGH (END COMMUNICATION)
   return ReadSPI1();
}

This works fine (at least it seems to work). Code runs ok and O'scope shows CS pin going low, then 2 groups of 8 pulses in the SCK pin followed by CS going high.

Then I have this:

unsigned char NRF_PutTXPayload(unsigned char *data, char len)
{

   if (len > TX_DATA_SIZE)
       return 0;
   SPI1_BEGIN();

   //SEND BYTE 1
   IFS0bits.SPI1IF = 0;
   while (SPI1STATbits.SPITBF);
   WriteSPI1(0xA0);
   while (!IFS0bits.SPI1IF);

   //SEND BYTE 2
   IFS0bits.SPI1IF = 0;
   while (SPI1STATbits.SPITBF);
   WriteSPI1(0xEF);
   while (!IFS0bits.SPI1IF);

   //SEND BYTE 3
   IFS0bits.SPI1IF = 0;
   while (SPI1STATbits.SPITBF);
   WriteSPI1(0xEF);
   while (!IFS0bits.SPI1IF);

  SPI1_END();
  return ReadSPI1();

}

As you can see this function above is a modified version of the original one I wanted to use. I'm currently not using those arguments and I used to have a For loop to send variable length group of bytes. Then I came into to this version above because it was freezing my PIC somewhere inside the For loop. Then I realize that if I keep only the first two blocks (SEND BYTE 1 and SEND BYTE 2) it's ok. When I add the third one, it freezes the PIC.

I know that I could reduce my code when taking care of flag bits but I copied exaclty the same lines from the WriteRegister function on purpose to make sure it should work.
Also, it should work because first I reset interrupt flag; then I wait for TX empty buffer to start copying; then I simply write to TX buf; and then I wait to interrupt flag which means transmission has finished. Altough not being the best/fastest way to do that, I think it should work well, but it does not.

It's strange because it works with 2 bytes but not with three. The O'scope shows CS pin going Low, then it shows the three groups of 8 clocks in SCK pin but the CS pin never goes back high (and the PIC is freezed).

Any idea what might be happening?