Your findings seem to be correct
Since time is a actual decimal I did this:
CDbl([end_time])*10000000000000-5416666666666
To find the xth decimal point of this value to be: 0.6669921875
CDbl(#1:00:00 PM#)*10000000000000-5416666666666
To find the xth deciaml value of exactly 1pm: 0.666015625
13/24*10000000000000-5416666666666
The part representative of 13 hours on 24: 0.666015625
I cannot seem to get a reliable 1:00:01 PM time stamp for some reason but yes, it seems like your value is slightly larger than 1:00:00 PM
And yes it is a date/time field, with only dates and only times, however if you use TimeValue([end_time]), it seems to get truncated to actual seconds and the field "reverts" back to =1:00:00 PM
([End_time]-TimeValue([end_time]))*10000000000000000
This returns a difference of 1.11022302462516
There is another "end time" at id 1865, which has the exact same deviation but minus instead of positive? At a time of 8:45:00 PM, there has to be a cominality between the two?
This is a total first for me to ever see this, seems like (perhaps) A2007 is able to record milli seconds like SQL Server? Or some other "anomily" going on?