It's unlikely that I can qualify for the job of "enlightening" you, since that would probably require me to become enlightened myself. However, I can attest to the fact that Microsoft has published a lot of theoretical information and some empirical data about the optimization of file allocation unit sizes. It's fairly obvious that they at least believe their own data since they did try to enable the new CONVERT utility in WinXP to produce 4,096 byte clusters whenever possible. (It's not always possible because the demarcation between partitions may not occur on the requisite multiple of 512 byte hard drive sectors.) As it happens, 4K is a number of considerable significance to other parts of NT-based operating systems than the file systems. It is the optimal (or only) size for chunks of data used by any number of entities in the OS. It also is supposed to be the optimum balance between tendency toward file system fragmentation (which is made worse as cluster size decreases) and file system slack (which is made worse as cluster size increases). I presume that these considerations have something to do with the choice of 4,096 bytes as the point above which encryption and compression cease to be available in NTFS as well.
I haven't read the FAQs on choosing between the FAT32 and NTFS file systems here, so I'm not sure they address this particular issue, but they were written by some people who seem to be pretty sharp traders -- so I imagine they might be worth a look-see. And, of course, there's always the MSKB, MSDN and other arms of the mighty Microsoft online presence -- when you have a little time.
I hope this is a little bit helpful. You'll have to consult more authoritative sources if blinding light is what you seek! In any event, I think that a partition using 512 byte clusters would probably not prove to be a terrible stumbling block to most of us -- unless the partition happened to be exceptionally large and / or was allowed to become very badly fragmented.
- Collin