Not quite true. Inserting a row into a clustered index never causes the entire table to be reordered, it simply maintains the page pointers on split pages so that the page chain remains in sorted order. What it does not do is attemtpt to maintain the pages in contiguous physical order on disk. So, suppose there are two pages, AA and BB, and the page chains on each point to the other, forward and back (AA BB). Now we insert a row that falls onto page AA, but AA does not have room for it. SQL will do a page split by grabbing a fresh page (lets call it AB), copying half the date from AA onto AB, and updating the page chain so that the data is in order as the chain is walked (AA AB BB). These three pages do NOT exist contiguously on disk. In doing the page split, pages before AA (00 - 99) and after BB (CC - ZZ) are un affected.
One big advantage of clustered indexes that seems to be under-emphasized in these posts is that at the leaf level of a clustered index you find the entire data row. Therefore, SQL's optimizer will favor a clustered index, knowing it will find your data there. With a non-clustered index, the leaf level holds an address where data can be found, and SQL then has to go to that address to get the data. Metaphorically, a clustered index is like an encyclopedia, and a nonclustered index is like a set of textbooks. If you want to look up "Gettysburg", grab the G volume of the encyclopedia, and rapidly flip the pages to find the entry. Further, not only did you find "Gettysburg Address", but "Gettysburg, Battle of" is there, too. In the textbook, you can rapidly scan the index for Gettysburg, find out it is on page 429 and 814, then turn to those pages to find your info. The non-clustered index beats the heck out of reading the whole book, but is not as efficient as the clustered index.
As a rule of thumb (and if that's not strong enough, throw in a few fingers, too), EVERY table should have a clustered index. The trick is in identifying which index it should be. In all the years I've been a DBA (12, gulp!), I can only think of one time when I made an argument to uncluster a table, but that's a story for another day.