Did you do a manual count? The computer could have certainly helped you with the word count: the French text has 62 words, the machine translation 76, Human Translator A 63 and Human Translator B 59.
The original French appears to have 43 types to its 62 tokens;
The machine translation has 40 types to its 76 tokens;
Human Translator A has 41 types to 63 tokens;
Human Translator B has 48 types to 59 tokens.
Even these very rough data show that the machine translation repeats the same words again and again, and that when we compare Human Translators A and B we see the latter not only using a smaller number of words overall but a higher proportion of different words, outdoing the original French both in economy and in breadth of vocabulary.
The counting may not have been as straightforward as you imagined. You may have wondered if ‘twice-yearly’ should be counted as one or as two words or whether, in the French, you needed to count ‘d’’ as separate from ‘de’ or ‘l’’ as either ‘le’ or ‘la’.
Also, in comparing overall numbers of words in an original and its translation one would have to take into account the structure of the respective languages and conventions governing genres and text types.