Welcome to BLF! 
You are correct; a 4x7135 light at 100% should be more efficient than a 8x7135 light running at 50. So, the half-power modes on this are less efficient than a full-power mode on a driver with half the chips removed. But the difference isn’t very large. According to djozz’s measurements, a XP-L V6 emitter should produce about 700 lumens at 1.5 amps or about 1275 lumens at 3 amps. The circuit only goes on and off, no intermediate levels, so half power means flashing on and off very quickly and it doesn’t really increase efficiency. That means the 50 mode should do 50% of 1275, or ~638 lumens. As compared to 700 lumens for a native 1.5 amp driver.
The runtime for both should be about the same, but the half-of-3-amps driver would perhaps be about 9% dimmer. Human eyes can’t generally see a difference of 9% though.
The point of the half-power mode groups isn’t efficiency; it’s convenience. It provides an easy way to limit the maximum power without changing any hardware.
The new firmware may increase runtime by offering a few lower modes (like 1% instead of 5%), but it won’t change the runtime for anything running at the same power level as an older driver. One amp is one amp, regardless of what’s being done with it.
Long story short, the 4-chip driver should be about 10% brighter than the 8-chip driver at half-power, but the runtime should be the same. In person it would be hard to tell the difference, even with both lights side-by-side.