Yes, Fig. 1 is not completely clear. I think m should be the raw number of training examples, to be consistent with the notation. I checked http://www1.ics.uci.edu/~mlearn/MLSummary.html, where one can find that for "lymphography" there are 148 examples, which is consistent with Fig. 1, but "adult" has 48,842 examples (some with missing attributes).