I have worked with large data warehouse applications that do a lot of data aggregation. The ability to roll data up into non-day increments (hours, months, quarters and years) is useful in those applications. It allows you to store the aggregated data referenced by a single time coordinate. I'm not suggesting that a set of data be linked to ALL dimensions. So you wouldn't link a record to the year 2008, Quarter 1, Month 10 and day 10 if it was 10-OCT-2008. But if you do periodical aggregation as the need for granularity decreases as data ages, the ability to use a single value to represent that can be useful.