Suppose you have a disk that goes around π*e times per
second, i.e. about 8.54 times per second. You could
record every time it comes around, but suppose you
don't want to. Suppose you prefer to have data points
spaced *approximately* one second apart. So wait until
the beginning of the next whole second, then timestamp
the next event (i.e. the start of the next cycle).
Keep a bazillion guard digits on the timestamp.
Using this procedure, the cycle numbers are exact,
and the timestamps have negligible roundoff error.
The spacing between timestamps is not quite even,
but is as even as it can be without damaging the
data.
If you absolutely insist on data that is evenly spaced in
time, start with the timestamped data in all its glory, then
fit it to a physically-correct model, then use the model
to generate whatever you want.
The Vernier instrument presumably tries to interpolate,
but it cannot possibly do so properly, because it has no
idea how to model your data. Any interpolation scheme
that works OK for your situation will fail miserably in
somebody else's situation, and vice versa.