Dear Mattias,
no that is not the problem, I believe. I think that Mathematica treats your equation quite correctly. The solution looks rather straight because of your parameters. If you use different parameters:
Equation1 =
NDSolve[{L''[
t] == (((0.1*4 ?*10^-1*
L[t]*(440*Exp[1]^(-0.001 t/(0.5*0.0025)))^2/
?)*(0.01^2 + (L[t]^2 - L[t]*Sqrt[0.01^2 + L[t]^2]))/
Sqrt[0.01^2 + L[t]^2])/0.01), L[0] == 0.1, L'[0] == 0.01},
L, {t, 0, 0.1}]
looks not that linear anymore:

I think that for your parameters the straight line is a good approximation in your domain, but the result is strictly speaking not a straight line.
Cheers,
Marco