As was just said, it's frequently quicker to let SQL index the entire table once the data is in it, building just the index pages, than to make it write a record to a data page and then turn around and have to find and update an index page as part of the insert.

With a clustered index, on the other hand, the data is being written in the same set of pages as the index, so it would be more efficient (and this is an educated guess, I don't have benchmarks to verify this) to have the clustered index in place when inserting data. And if the data was being inserted in the same order it will be physically stored in the clustered index, so much the better.

Maybe others can verify this, or correct my thinking....