In fact, the answer to your question here is "not exactly", and your followup statement about max power consumption is incorrect. That's because chips don't run at their maximum possible speed in actual devices.
M chips run at lower speeds in iPads than in MacBooks, specifically due to heat (and to some extent also battery) considerations. They will ramp down in speed further under heavy use to keep thermal load constrained. Apple carefully controls this in software. People sometimes refer to this as throttling, which is a sorta-accurate but misleading word.
The right way to think about it is this: Apple have a thermal and power use ceiling they don't want to exceed. And thanks to physics, those two factors come out to basically the same thing: thermals and power use will move as one. (Well, not exactly, but they head in the same direction, so you ratchet up allowed performance until you hit whichever limit you hit first.) They limited the M2 in the iPP to that chosen level. You can expect that they will do the same thing with the M4. So, while my "for the same performance" isn't directly relevant to the iPad, "for the same heat" is. "For the same battery consumption" is too, mostly, though it may change a bit based on changes in the size of the battery. And those again strictly favor the M4 over the M3 and the M3 over the M2, meaning my original conclusion is still correct.
It is likely that the new copper and graphite heat spreaders in the M4 iPPs give Apple the leeway to allow the M4 to run hotter than the M2, actually, which skews things a bit. But probably not enough to change any conclusions, because they still face battery limits even if thermal headroom has improved a bit.
Coming back to the original point: Because chips can run at a wide variety of clock speeds depending on thermal and power limits, and do so in different devices in the real world, you need to control for that to talk about them meaningfully. So, your "0-5%" comment is true, and it's useful in the context of comparing M2 to M3 MacBooks. But it's not useful for talking about expected behaviors in a different device like an iPad.