I looked at the I0QM 1:1 design. The voltages at the load are very different and the mismatch bandwidth is not that great. It may have an application, but I doubt it is suitable for UHF TV.
A friend just modeled a quarterwave 75-ohm line in series with a quarterwave 90-ohm line, simulating the sudden change in trace width and characteristic impedance of the UHF-TV printed circuit balun whose image I posted. It turns out to match about 250 ohms instead of 300 ohms. I tried reversing the order of the lines and it matched 350 ohms! This not-exactly-4:1 impedance transformation could be quite useful.
For a 4:1, the optimal impedance is 150 ohms, which could be implemented on a PC board. The resulting mismatch loss is negligible over the entire UHF-TV band.