The main reasons for the 1v1 testing is to identify a unit's cost effectiveness which is very important in MP, and see how strong the rock, paper, scissors system is. It's only a first small step to learning how to play a full scale battle. I would say it jump starts you on the MP learning curve.

Of course, you can do all your learning by way of battles, but you are going to loose a lot until you figure out what the most cost effective units are. Once you understand each unit's capabilities, you can concentrate on tactics. Some people pick up on the unit properties very quickly just by playing battles. It helps if your opponent analyzes the battles with you afterwards. Once the replays and logfiles are working, you'll be able to analyze your MP battles offline, and that should be of tremendous help to players working on improving their game.

The only other reason I can think of for doing 1v1 testing is if you want to change a unit's stats. You have to test the change to be sure that's it's having the effect you want. You're right that 1v1 tests give a false impression of a unit's resilience to routing. A unit that looks resilient in a 1v1 test might rout quickly in a full scale battle where there is a complex interaction of morale penalties. A ranged unit that looks fragile in 1v1 might stand too long in a full scale battle because it's usually positioned just in front of other units where it picks up morale bonuses. So, the way a unit works in full scale battles always has to be the final test, but it's also the most difficult thing to evaluate because of the interaction of so many factors including the differences the players themselve inject.