Indeed the issue that trips up many of the people I've worked with is that CLUSTER will make a table seem like the clustered index of other systems (in as much as it physically orders the table), but without an appropriately large fillfactor (and sometimes not even then), new rows and updates are not stored in indexes physical ordering.

So at first glance it's super-speedy for clustered index-like queries, but then becomes slower and slower over time as the data changes and consecutive index reads are no longer sequential on disk.