You have to be careful when doing differential measurements (x-y) with the iNVERT function, there is no standard for where the summing takes place. Many (all?) cheap LCD scopes just subtract the pixels, so although what you see is what you get, if one of the signals clips the screen edge the diff signal is meaningless.
Decent analog scopes (and some were quite cheap also) used to let you have seriously high gains which would overload each channel on their own, but when summed, cancelled the inputs so that you could get a good high-resolution view of a small signal which was superimposed on a large one (like a 100mVp-p crosstalk transient on a 24V DC supply).
Its worth checking this on any scope purchase. Of course, you will eventually build a decent diff pre-amp yourself, but until then, a good ADD-plus-INVERT on your scope is very handy.
That's kind of an age-old question. It's not JUST about accuracy. It's also about whether you change the measurement appreciably by just touching probes to the circuit.
IMO, every EE who's actually an EE has had this drilled into him throughout school and career, and has been bitten when he ignored the issue.
Sometimes the absolute number isn't all that important, but what always matters is whether the scope, voltmeter, ammeter, whatever, is making the measurements irrelevant.
Sort of a macro version of Heisenberg's uncertainty principle.