Uwe,
Discussions in the context of measurement are a different animal than what Max was asking.
When looking at the phase behavior of a device under test (DUT), the total phase is the combination of the DUT's inherent phase response, and the phase that accrues due to time of flight from the DUT to the measuring transducer.
In the naive case, the arrival time alignment delay of the two signals that are used to calculate the transfer function (i.e. the input stimulus and the measured response) can only be controlled by buffering on the interval of one sample. This means that the phase calculated by the TF could be off due to the limited granularity of the delay time you can pick. Note that this isn't a limitation of the math, but rather of one's ability to pick the alignment delay time for the signals fed to the transfer function. If you could pick the delay time precisely, the phase would be known precisely.
There are more advanced techniques available that allow for inter-sample delay time choices, FWIW.
