Thanks to all, I think I'm starting to understand how that Accuracy Table works: When there's an entry, say,
2% + 40 (digits)
then it means that the reading can deviate from reality by 2% of the true value (not 2% of the range) plus 40 times the resolution (i.e., whatever-the-smallest-unit-on-that-scale-is). Therefore, if the true value is 0, then 2% of that is also 0, but 40 digits means 40 times 0.1 Volt (if 0.1 Volt is the resolution) ergo 4 Volt.
Hmm. If this is true, then the Fluke 289 (with shorted leads) should also not give a reading of more than 0.06 mV when set to AC mV, since the resolution in that setting is 0.001 mV and the accuracy is (at worst) 1.5% + 60.
Right?