If claim 1 is tossed then subsequent claims based on claim 1, which will probably be many, are automatically tossed. Therefore claim 1 better be your best claim.
For example, in "Method and system for mapping between logical data and physical data," if Claim 1 is tossed, then claims 2-9 automatically go with it.
Claim 1. A method for surviving, comprising: Cooking a meal.
Claim 2. A method as in claim 1, wherein the meal is breakfast.
Claim 3. A method as in claim 2, wherein the breakfast comprises sausages and bacon.
Claim 4. A method as in claim 3, wherein the sausages and bacon are fried.
and so on. A court might decide that there's some prior art for having meals, and toss out claim 1. But that prior art might only be for lunch and dinner, in which case claim 2 might survive. On the other hand, maybe someone else has had breakfast before (publicly enough to count as prior art). Then claim 2 goes too, but maybe claim 3 survives. Oh, too bad, it turns out that the court thinks having sausage and bacon for breakfast is obvious to one ordinarily skilled in the art. Bye-bye to claim 3, but maybe frying them is an extraordinary new idea that no one else would have thought of.
Of course, similar things can happen during patent prosecution (i.e., filing; the terminology is silly), only now it's the patent office that keeps objecting to your claims and requiring you to fall back to weaker dependent claims.
It's absolutely standard for your first claim to be very broad, and subsequent ones to be narrower. It's absolutely not standard for claim 1 to be "best" in the sense of "least likely to be invalidated".
Yes, but if you're not doing Claim 1 ... i.e. you're not cooking a meal ... even if your meal is breakfast, and the breakfast comprises sausages and bacon and those are fried ... you're not infringing.
Salesforce can defend two ways: 1) challenge the validity of the patent, or 2) prove that they don't infringe the patent.
Option 1 is harder. Patents are presumed valid. Defendant (Salesforce) has the burden of proof that the patent should be overturned. Salesforce will have to do that for each alleged infringing claim.
Option 2 is easier. Plaintiff (Microsoft) has the burden of proving infringement. Salesforce just needs to show that they aren't infringing claim 1 and then they also aren't infringing claims 2-9.
Think of your quotation as a claim, that describes the set of all possible inventions that meet that claim. Within that set are all kinds of mappings, between different kinds of logical data and kinds of physical data, in different situations, and using different methods for doing so. Subsequent claims restrict that set, until (the inventor hopes) you get one that is novel and inventive etc enough.
Perhaps you're thinking of the term "dependent claims", which describes claims 2-9 in your example. They're not dependent in that way.
For example, in "Method and system for mapping between logical data and physical data," if Claim 1 is tossed, then claims 2-9 automatically go with it.