We do not have $\mathcal L_X Y = D_X Y$. The actual statement is $\mathcal L_X Y = D_X Y - D_Y X$ (which, according to some authors, is a definition), where this is defined in charts. This ends up not depending on the choice of chart; but $D_X Y$ does depend on the choice of chart.
The Lie derivative $\mathcal L_X$ makes sense as something that acts on a lot more than just vector fields ($L_X T$ makes sense for $T$ a tensor - so in particular differential forms or even functions). In the case of functions, $$(\mathcal L_X f)(x) = \frac{d}{dt} f(\phi^t(x))\bigg|_{t=0},$$ where $\phi^t$ is the flow of $X$ at time $t$. Because $\gamma(t) = \phi^t(x)$ is the integral curve of $X$ through $x$, by definition this is $X_x f$; so $\mathcal L_X f = Xf$ (which is sometimes denoted $D_X$). This is a context in which "$\mathcal L_X = D_X$" is true; for higher-order tensors the formulae, as above, are more complicated.