When transactions are supported, while it is true that every update BELONGS TO a transaction, it is not true that every insertion IS a transaction.
I'm nitpicking, but with Oracle, DB2, Sybase, etc. every DML operation is run in a transaction. The transaction may automatically commit at the end of the operation (i.e. if AutoCommit is on), but there is always at least an implicit transaction as even a single row insert may in fact generate more operations through triggers (inserts to a shadow table, updates of summary tables, etc.) and these all will be guaranteed to perform as a single operation, even in the absence of explicit transactions in the DML.
That being said I think that keeping the transaction logic out of the examples is a good thing, as long as their use and functionality is explained somewhere in the document - after all transactions are pretty central to RDBMS systems...