(Answer, take 2)
You can get a reasonable approximation of the phase with
f[tt_?NumericQ] := -(ArcTan @@ ({x[t], x'[t]}/
Sqrt[x[t]^2 + x'[t]^2]) /. s[[1]]) /. t -> tt
Here are a few stories. First we show driving and result together. This means that they are a little out of phase.
plotx2 = Plot[
Evaluate[{x[t], Cos[t - 0.5*0.0000652*t^2]/5} /. s], {t, 0, 100},
Frame -> {True, True, False, False}, FrameLabel -> {"t", "x"}]

Now we show the two phases together. This time I draw several times.
phaseangles =
Plot[{f[t], Mod[t - 0.5*0.0000652*t^2, 2*Pi, -Pi]}, {t, 100, 120},
Frame -> {True, True, False, False}, FrameLabel -> {"t", "x"}]

Recently, phase differences have been shown.
phasediffs =
Plot[{f[t] - Mod[t - 0.5*0.0000652*t^2, 2*Pi, -Pi]}, {t, 100, 120},
Frame -> {True, True, False, False}, FrameLabel -> {"t", "x"}]

Maybe I have something extra (the terms Mod [] become intrusive), but this should give an idea of how to proceed.
Daniel Lichtblow Wolfram Research