I'll give this a go ...
I think your second method is better than your first method but applying the mortality rates to the populations at t1 and t2 (as you do) will probably overestimate the expected numbers of deaths at t1 and t2 and the difference between them.
I think we need to assume that there are t0, t1, and t2 which are the same distance apart. The first mortality rate (0.021) applies to the time between t0 and t1. The second mortality rate (0.022) applies to the time between t1 and t2.
We should not treat deaths as simultaneous events at which occur only at points t1 and t2. We usually have no information about when each death occurs so we assume (i) deaths can occur at any time point between the two time points, and (ii) deaths occur more or less evenly between two time points. This leads us to applying the rate to the population at the mid-point between the two time points.
Using your numbers ...
We know:
t1 population = 50000
t2 population = 53000
Population growth between t1 and t2 is:
(53000 - 50000) / 50000 = 0.06 (6%)
If we assume constant growth then:
t0 population = 50000 / (1 + 0.06) = 47170
We have:
mid-point population t0:t1 = (47170 + 50000) / 2 = 48585
mid-point population t1:t2 = (50000 + 53000) / 2 = 51500
This gives:
deaths t0:t1 = 0.021 * 48585 = 1020
deaths t1:t2 = 0.022 * 51500 = 1133
The difference is:
d = 1133 - 1020 = 113
I am not sure that I would call these "excess deaths" since mortality of t0:t1 and t1:t2 are almost identical. In this case the additional death are are due to population growth not to changes in mortality.
As for a good book ... this topic is covered in most introductory epidemiology and biostatistics texts.
I hope this is of some use.