I was looking at my completed models just now and while credits for all of short tasks completed are the same, the cpu and run times vary widely! The extremes being
http://climateapps2.oerc.ox.ac.uk/cpdnboinc/result.php?resultid=18402721 at 298.71 according to the table and
http://climateapps2.oerc.ox.ac.uk/cpdnboinc/result.php?resultid=18448967 at the other extreme is shown as a cpu time of 155,529.80

The full table of recently completed tasks can be seen here.
http://climateapps2.oerc.ox.ac.uk/cpdnboinc/results.php?hostid=1354308&offset=0&show_names=1&state=2

Strangely, the seconds/timestep show very little difference.

My assumption is that the cpu time taken is being misreported. However I have no theory as to why this might be as of yet?

It looks like the completed models that have the fewest CPU sec on the task listing are those that ran completely without a quit request.

Looking at the trickle listing on the completed tasks, they all have similar CPU times listed there.

This underreporting of CPU time was reported on the moderator mailing list back in August of last year. It had something to do with the boinc build used in compiling compound apps and a bug in that boinc build. Thyme and Richard H were posting about it. It's a cosmetic error as it doesn't seem to affect the results or anything else other than the CPU time in a listing of tasks.

Thanks for that. As the times to complete on my computer hadn't shown anything alarming I assumed it was a misreporting. My only just noticing it probably says something about my lack of observations skills!