With most loadtests, it's assumed that you should start with the first row in the data-table, and that the system under test has been restored to a 'clean' state.
When developing the test, and if it's a pain to clear out created objects (such as users) I usually do the following
1) save my entire datatable as a seperate XLS file
2) Nuke rows from the data-table as needed to make sure it always consists of only 'fresh' data.. e.g. if I was debugging and got through 2 iterations, then I'd nuke the first two rows...
Note, when multiple vusers are used, and a set number of iterations is specified, the table is usually allocated a vuser at a time. e.g. if 3 vusers running 4 iterations each, then rows 1-4 are set for vuser 1, rows 5-8 to vuser 2, etc.
If the script in that example runs and bombs during first iteration, then you'll need to go in and delete rows (from the bottom up if you are smart) 9,5,1 to remove the 'used' or 'dirty' data.
(or scrub the system under test back to it's starting config)
A project without change control is by definition out of control