Then you've got the very first case I've ever seen where "clock format" calculates the value incorrectly.
Just for fun, though; Could you post the output of this:
As you can see the unixtime is correct and your test are correct, but of some reason the clock format shows totaly wrong time and now is the diff almost a minute
clock format does show the correct formatting for the given timestamp(s). I do notice that you are mixing unixtime and clock seconds as the source for the "unixtime", which would suggest that these are returning different values at any specific point in time.
The unixtime command relies on the time(2) system call, while clock seconds uses the gettimeofday(2) system call. Both of these should return the value of the system clock, thus I have no further ideas of the issue right now.