Tune Tx for maximum smoke LOL! This occurred when your anode choke was a bit under-specified for the PA voltage!
ON HF it was easy because we didn't even try to get a 50 ohm match. Trying to get a match to a 50ohm feeder only became important at VHF and above.
So the technique on HF was to use a transmitter with Pi tank comprising a tuning capacitor and a loading capacitor. With loading capacitor at maximum capacitance, you tune the tuning capacitor for the bottom of the dip in PA supply current. Then you slowly reduce the loading capacitance while retuning for the dip, which gets shallower. You have got it right when the dip is shallowest but still discernible.
I finally understood what was going on. The driven PA generates an RF anode (or collector or drain) current. When the tank is tuned to resonance, this develops a voltage in-phase with the current so that maximum current occurs at minimum PA anode (collector or drain) voltage, minimizing average DC consumption and causing the current dip.
What you want to then do is to increase the load on the output (by reducing the loading capacitance) so that the anode/collector/drain still swings as near as possible to 0 volts, but not so much that the anode/collector/drain voltage doesn't achieve the full voltage swing. When the latter happens, your power output is current limited and the dip won't happen. So it is when the dip just happens that you have it loaded just right.