To my way of thinking - with the top method, the 470R acts not only as a screen grid resistor, but as an unbypassed screen supply resistor. This means that the gain of the stage will otherwise be less than if you had the resistor bypassed, because of the action of the screen current feedback in varying the screen voltage. If you don't like the effect of the 470R, then try upping it to 1k, 1k5 or even up to 4k7 which would make the attack softer and squish the grid curves together more. But there may not be enough filtering (at the B+ node) for the screens this way
But the next method down will have better filtering (but even better filtering still, if the 1k was replaced with a 3-5H 90mA choke) and more gain AFAICT, because more of the screen supply voltage is being clamped.