I am not able to see any visible improvements due to the BURST transfers. The same BURST transfer could be done through multiple SINGLE transfers, in the same number of cycles. Then what is te advantage of having BURST transfers?
HBURST can be used by a master to indicate exactly how many transfers it wishes to perform (4, 8 or 16)
This can then be used by an arbiter to decide when it should change the granted master (usually more efficient if you allow a master to complete a known length transfer sequence).
It can also be used by slaves that can optimise their responses if they know in advance how many data transfers there will be. The typical example here is a DRAM which has a high initial access latency, but sequential accesses after that can be performed with fewer wait states.