YJ's models convert either grid bias or cathode bias amps to use EL84's in cathode bias.
Cathode bias push pull.... is really a poor design. Here's why -
Normally push pull amps have a set of components that make up whats called a "phase inverter", whereby, the signal from the pre-amp is split into two signals, one of which is upside down from the other.
The two split signals allow push-pull operation, whereby, one signal is fed into one output tube, and the upside down signal is fed into the other tube. The output transformer contains back to back windings that are opposite each other (really just a coil tap...). This allows the output tubes to be set to conduct, not the entire wave of signal, but only part of it. So, each output tube is really only conducting part of the time, and resting in between conduction. This resting period allows things to cool, and allows greater signal strength to occur when the tube is conducting. So each tube conducts more signal strength, for less time, and by doing that, they dont overheat. When the two signals are fed into the back to back windings, the combined signal on the speaker side of the output transformer is complete.
This resting, cooling cycle is one way to think of the operation class of the amplifier. Class A = 100 percent signal conduction. The tube is "on" all the time. Class B = 50 percent duty cycle. On exactly half the time, off exactly half the time. Class A/B is someplace between the two Classes, with mostly on, and some off duty time.
The problem: When the tube conducts all the time, in Class A, its current use is just about constant. For example, a Fender Champ might use 45milliamps resting, and only 48-50milliamps when dimed and playing power chords. Champs are Class A amps and that 6V6 tube they got takes a lot of heat because of it. In a Class A/B amp, the current use goes up, as the signal strength goes up. So in a Fender Princeton Reverb, those same 6V6's only use 25milliamps resting, and about 50milliamps during dimed power chords.
So whats the problem? It has to do with that self bias thing. You know I was getting to that part....
In a Class A amp, its easy to "bias" the tube (set its overall operating level) by using whats known as a cathode resistor which has the effect of creating a voltage offset (ie, bias=offset) at the grid of the tube, setting its level. Easy and simple to do, and sort of self adjusting too. Since the current use hardly changes, the voltage created across the resistor stays very even, a good thing.
In a Class A/B amp, the output tubes are normally provided with a fixed voltage at the grids of the tubes. Thats because if a cathode resistor was used, the voltage created across the resistor would change as the current flowing through the resistor changed. So if we went from 25milliamps to 50milliamps, the voltage would double... not good. Hence the "fixed" bias (aka grid bias).
Converting an amp from push pull A/B to push pull A operation would require careful calculation or the change of the output transformer. What the Y/J's do is fake it. They go Class A at low volume, and slip into A/B at higher volume - because the current use changes and the bias level therefore changes as well.
Having said all that -
Matched tubes are a must, because although self biasing, you want both tubes to track each other closely in duty cycle as you go from low to high volume. You'd get some UNpleasant distortion, quite possibly, if you didn't do that.
So yes, match the tubes... closely even.