Allow me to share with you a little snippet of Atmel's flash API example code. Notice not only the write to DPTR but also the even more hair-raising use of the accumulator. I'm not sure why they didn't have the balls to write directly to R1 as well:

Uchar __api_wr_code_byte (Uint16 address, Uchar value) small
{
bit ea_save;

ea_save = EA;
EA = 0;
DPTR = address;
ACC = 0x02;
ASM_MOV_R1_A();
ACC = value;
MAP_BOOT;
__API_FLASH_ENTRY_POINT();
UNMAP_BOOT;
EA = ea_save; // restore interrupt state
return (ACC);

}