NavList:

"Rounding bias", the computer arithmetic people call it.

When glancing at the watch, it might be possible to remember whether the second changed immediately after the glance or somewhat later, in effect splitting the second with the eye/brain.

Cheers,

Peter

ps: To take this to slightly absurd lengths:  the statistics of this kind of error is a little different---the time error is uniformly distributed over one second rather than a more usual, softer, Gaussian-like distribution.  A navigation filter might want to take that into account.  For example, suppose you have many sights all around the horizon.  The LOPs will form a circle around the AP, with some fuzz.  But the fuzz will have hard edges (in the longitude direction) because the time error is sharply bounded from 0 to 1 second.