Your first example: whenever I have had to interface data with another company, the only thing I have had to worry about is the format of the data. The languages were not important. The only reason I can think of for the new company to have to switch to COBOL would be if they were bought out by the other company and forced to work on the same machine. Even when I've seen this happen (in my last company, we had four states merge into one), it has still been a matter of accepting a standard for the format and sticking to it (I wish Micro$oft understood this). Then, when we developed new systems, they often were not in COBOL (unless we had a dinosaur leading the team).

As for the second example, that one is trickier. Maybe my challenge to you should have specified "a new company that's not performing mouth-to-mouth on a stegosaurus". I may be forced to sing your praises simply because I worded my "deal" wrong.

In any event, I see that both of your examples are dependant upon current implementations of COBOL. If a company is starting from scratch, why would they use it?