Okay, I am having trouble understanding the why of your reference code as well, so here is what I did:

SELECT run_time

, run_time/10000 AS run_time_hours

, (run_time%10000)/100 AS run_time_minutes

, (run_time%10000)%100 AS run_time_seconds

, (run_time/10000 /*run_time_hours*/ * 60 * 60 /* hours to minutes to seconds*/)

+ ((run_time%10000)/100 /* run_time_minutes */ * 60 /* minutes to seconds */ )

+ (run_time%10000)%100 AS run_time_elapsed_seconds

, CONVERT(DATETIME, RTRIM(run_date)) AS Start_Date

, CONVERT(DATETIME, RTRIM(run_date)) +

((run_time/10000 * 3600) + ((run_time%10000)/100*60) + (run_time%10000)%100 /*run_time_elapsed_seconds*/) / (23.999999*3600 /* seconds in a day*/) AS Start_DateTime

, ((run_duration/10000 * 3600) + ((run_duration%10000)/100*60) + (run_duration%10000)%100 /*run_duration_elapsed_seconds*/)

, CONVERT(DATETIME, RTRIM(run_date)) + ((run_time/10000 * 3600) + ((run_time%10000)/100*60) + (run_time%10000)%100) / (86399.9964 /* Start Date Time */)

+ ((run_duration/10000 * 3600) + ((run_duration%10000)/100*60) + (run_duration%10000)%100 /*run_duration_elapsed_seconds*/) / (86399.9964 /* seconds in a day*/) AS End_DateTime

FROM msdb.dbo.sysjobhistory

Hopefully this shows my work.

FYI, the 23.999999 is required to get to the correct precision of seconds.

After doing this, I am getting a factor of 2.777E-4 versus a factor of 2.893E-4 in your current code. This seems to me to be a difference of rounding. The one thing worth noting is that my solution involves taking the modulus of a modulus, versus the other solution which does not.

While I know this doesn't answer your question, I hope this helps get to an efficient solution!