Because the Romeo figure is a sample, not the entire population. One
normally makes population estimates from samples.
> Isnt it a bit misleading to state that 55% of the journals allow either
> pre- or post print self archving when the actual number is closer to 15%
> (24000/3897)? (irrespective of whether one believes that publisher's
> consent to self archiving is necessary or not)
No, it is quite reasonable to state that 55% of the journals sampled
support self-archiving, and to extrapolate from the sample to the
population as a whole. (The only exception would be if there was any
reason to believe the sample was unrepresentative or biassed. But since
the Romeo sample was of the top journals rather than a random sample,
any bias from this is likely to be toward an under-estimate rather
than an over-estimate of the number of journals already formally
supporting self-archiving.)