The decreasing assumption is not needed, as mentioned by Connor Harris in the comments. I claim that for $a < b < c$ we have
$$\frac{f(b) - f(a)}{b-a} \le \frac{f(c) - f(a)}{c-a} \le \frac{f(c) - f(b)}{c-b}.\tag{$*$}$$
Do you see how to obtain the answer to your original question from ($*$)?
Proof of ($*$):
By convexity,
$$f(b) \le \frac{b-a}{c-a} f(c) + \frac{c-b}{c-a} f(a).$$
This yields
$$f(b) - f(a) \le \left(\frac{b-a}{c-a} f(c) + \frac{c-b}{c-a} f(a)\right) - f(a) = \frac{b-a}{c-a} (f(c) - f(a))$$
and
$$f(c) - f(b) \ge f(c) - \left(\frac{b-a}{c-a} f(c) + \frac{c-b}{c-a}f(a)\right) = \frac{c-b}{c-a} (f(c) - f(a)).$$
To answer your original question:
If $t \le x_2 - x_1$ then $x_1 \le x_1 + t \le x_2 \le x_2 + t$. Using ($*$) twice yields $$\frac{f(x_1 + t) - f(x_1)}{t} \le \frac{f(x_2) - f(x_1 + t)}{x_2 - x_1 - t} \le \frac{f(x_2 + t) - f(x_2)}{t}.$$
Otherwise $t \ge x_2 - x_1$ which yields $x_1 \le x_2 \le x_1 + t \le x_2 + t$. Then applying ($*$) twice yields $$\frac{f(x_1 + t) - f(x_1)}{t} \le \frac{f(x_1 + t) - f(x_2)}{x_2 - x_1 - t} \le \frac{f(x_2 + t) - f(x_2)}{t}.$$