Data sheets are a good starting point, but in the world of reverse engineering, they don't rule. They can contain errors, and can occasionally deliberately mislead. I like to say that "truth is in silicon".

Part performance varies because of temperature, voltage, and process variations. With two of those numbers known, there is not all that much variance left with the process. That allows you to come up with a pretty good estimate of what the "true" spec is. And that number is what you use for your replacement chip design.

I totally agree that end users should be guided by the published specification. But many users will use a part beyond the published specifications. And so a successful replacement part needs to match the true specifications and not the published ones. Otherwise you will be overwhelmed by applications support to customers claiming that the old part works in their design, but yours doesn't.

The 82C55 contains one of the bigger mismatches between spec and silicon that I have seen.