The LDD (Load Indirect with Displacement) instruction is a variant of the LD (Load Indirect) instruction. It uses the Y register as a pointer and a six-bit displacement to offset the address. The displacement is a value from 0 to 63. I think what they are trying to say is that if you have a 'struct' of 64 bytes or fewer you can load any byte of the struct with one LDD instruction after loading the Y pointer only once.
This has nothing to do with the buffer arrays since the byte to be fetched is not known at compile time.
The 16-bit opcode is:
10q0 qq0d dddd 0qqq
d bits are the register to load
q bits are the displacement value