The z80's illegal interrupt mode

Hardware
1

If you look at Figure 4 of the z80 datasheet, you’ll see two interrupt mode flip-flops named IMFa and IMFb. Then this table is how to figure out what interrupt mode the z80 is in:

IMFa     IMFb
0        0        Interrupt Mode 0
0        1        Not used
1        0        Interrupt Mode 1
1        1        Interrupt Mode 2

There is no mode 3. Why didn’t they make the state where both flip-flops are set the unused mode instead? Surely they must have had a good reason to mess up counting in binary by skipping the IMFa=0,IMFb=1 state?

If we consider the names IMFa and IMFb to be simply swapped around and try to repair the table, then that makes the position of mode 1 make more sense when counting in binary, but not mode 2.

2

In think the answer is very likely to lie in the transistor-level implementation. If the circuit is simpler with some particular encoding, that’s a good encoding - leave it to the documentation to explain, if indeed the encoding is visible at all to the programmer.

So, there are several transistor-level resources for the Z80. Ken Shirriff has written some articles which are as usual excellent, but at present I don’t see anything about this.

But we do have some transistor-level simulation tools:
Visual Z80 on the Visual 6502 site
Visual Z80 Remix with improved facilities, also in-browser, by Floooh
Z80 Explorer (a Qt application for Windows and Linux) from Goran Devic

Video of Z80 Explorer: