There were 2 ways how USB keyboard and mouse support could be implemented for legacy software that pre-dated USB.
The 1st method actually made use of a real IBM 8042 compatible KBC. Most SuperIO chips always had KBC and some chipsets (VIA, NVIDIA) even have KBC integrated. So in this scenario, legacy software actually talked to KBC using port 60/64 and USB acted as a virtual wire periodically polling the data from USB keyboard/mouse and pumped the data into the KBC. From the KBC perspective, it was the same as receiving the data from PS/2 keyboard/mouse over wire. This was the most popular method simply because there wasn't a lot of saving the hardware manufacturers could bargain by cutting out the KBC. In this method, the typical BIOS option was USB legacy Enabled and Port 60h/64h Emulation Disabled. This worked for 99% of the time and the only time that one needed the additional Port 60h/64h emulation was for software that manipulated keyboard/mouse features directly such as using the NumLK, ScrollLK and CapLK LEDs as running lights. This was not a typical use case but there could be other industrial equipment that might require such to work properly.
The 2nd method was called the "virtual-KBC" implementation and it was typically touted for so-called "Legacy Free" PC, a new campaign for ditching all the legacy I/Os (serial, parallel ports etc.) and everything on USB. In this method, the entire IBM 8042 KBC was emulated by the USB software stack in the BIOS and hence Port 60h/64h Emulation became a must. This wasn't a very popular method because it was hard in the beginning to get the emulation as real as a HW KBC for timing sensitive software. At the same time, it was also an effort with a dead end because everyone knew OS without native USB support (such as DOS) was phasing out and UEFI would be the future.