IMHO, even with only 8-9% of the vote, the Lib Dems would get considerably more than 15 seats: nationwide polls never take into account the extent of the "local goodwill factor" to individual Lib Dem sitting MP's.
This is partly a long-term historical factor, since the Lib Dems have a considerable record (compared with the main parties) for local grass-roots campaigning, council seats and so on, and there are distinct pockets of Lib Dem support which traditionally defy national trends.
I suspect that this time there'll also be a much more recent, short-term effect (suggesting the same outcome) in that many Lib Dem sitting MP's have never before been "in the news" so much as at this election: the proportion of Lib Dem MP's who have held ministerial office at some point over the last five years is enormous, compared with those from the main parties at previous general elections. This can also only help them locally, on voting day, I think.
Regarding "likely seat numbers" for Lib Dems, Nationalists, UKIP and the Greens, I think Lord Ashcroft's local polling is likely to be far more reliably indicative than the nationwide polls now performed daily by the big polling companies.
Betfair's pre-election market - which of course takes updated polling information into account - suggests about 25/26 Lib Dem seats, rather than 15. On this subject, I trust that perspective a lot more than the Telegraph's.
They work out the number of seats from the poll by this,
Nothing to do with who done the poll. Ashcroft's polls would go through the same algorithm, his last poll has libdems on 9% also.