You significantly overestimate the losses of gravity drag and get some numbers wrong.
The first stage burned for about 2 minutes and 41 seconds or 161 seconds, not 304 seconds which you used for the burn time. Adjusting for this alone brings your solution to about 2284 $\frac m s$. But the Saturn V did about 2756 $\frac m s$ at first stage separation, so where does the rest come from?
$$
\Delta v=v_{\text{e}}\ln {\frac {m_{0}}{m_{f}}} \\
= 3000 \frac m s \cdot \ln \left ( \frac{2970t}{819t} \right ) \\ \approx 3864 \frac m s
$$
So we get about 3864 $\frac m s$ of $\Delta v$ from the first stage, and then we sub $161s \cdot \frac {9.81m}{s^2} \approx 1579$ to get $\approx 2284 \frac m s$
But this assumes the rocket went completely vertical, which is wrong.
Take a look at this graphic:

The point is you need to do vector addition of both the downward component (gravity) as well as the acceleration vector, which quickly becomes non-vertical. @RussellBorogove has given the graphic, from which you can see that the rocket is at abut 45° pitch angle at about 90s into the flight. Thus, gravity losses are minimized since the rocket actually doesn't point upwards.