I guess the person who made the question got tangled in his own definition. Indeed, if the "smallest addressable unit" is a "4 byte word", then it would follow that in 2^10 bytes there are 2^8 different addresses, which means that the highest address is 2^8-1 or 255.
In typical computer architectures though this is not the case. The smallest addressable unit is actually a byte (since all addresses represent the number of bytes from "the start"), but valid addresses are only those that divide by 4. So 0 is a valid address and 4 is a valid address, but trying to use 3 would result in an exception. This is called "alignment", by the way, and different instructions can require different alignment, which can vary anywhere from 1 to 128, and sometimes even more. Typically though it is a power of 2, because those are easy to work with.
I suppose the question was really about alignments, but the it got phrased in an awkward way. You can also try asking about this to your teacher, who should be able to clear things up.