Okay, normally we have exit polls - that's people collecting volunteered data at exit of the polling station. This then gives a national sample and reasonable results.
This time there were no exit polls (was looking for them and surprised they weren't there) so some groups went out onto the streets and did a poll there. This netted the figures people quote - that 62% of the 18-24's who voted, voted in (or 72% according to another poll), but the numbers sampled are low.
In the one case, 1,600(ish) people were polled and only those who were willing to answer were counted towards either vote. I can't remember if it was 1,900 or 19,000 for the other poll - the number was given, but it was (very) small print. Neither sets of figures seemed to add up, though, so I doubt they were really accurate or representative of how people did vote. For example, a straw poll here at work had more 'Out' votes from the younger workers and 'In' votes from the older, but that was an incredibly small sample size.