Can't you do both?
I've always been told to measure the temperature for the 1st, 2nd and 3rd minute.
Then mix the 2 substances at the 4th minute.
Then measure the temperature at minute-intervals, from the 5th minute, until you get 2 concordant results.
Then plot a graph of temperature vs. time and extrapolate both lines (i.e. the line before the 4th minute and the line after the 4th minute).
Therefore the change in temperature is the difference between the 2 extrapolated lines at the 4th minute.
I've been taught that this is one way of reducing the errors from the heat loss due to the apparatus.