Looking at those figures more closely, they do show a pattern under AV. As you might expect, the Libdems gain most 2nd prefs from both Labour and Tories, but that fact only has a small effect on the overall result – a handful of seats gained by the lds at the expense of the other two, but only a handful.
What seems, not surprisingly, to be the crucial figure is the 2nd prefs of the ld voters. I'm slightly surprised by the figures in that they suggest more 2nd prefs for the tories than labour right through the Thatcher years. That does surprise me, I admit, and it undermines some of what I've been saying. Then, there is a decisive shift towards Labour in 97, which is what leads to the collapse in the tory seats. Even relatively small shifts in the 2nd prefs of the libdems can have large effects on the results, it seems.
So, perhaps, AV would lead to a veering of strategists away from their obsession with wooing swing voters in marginal constituencies towards wooing LibDem voters in all constituencies. I agree that this is not exactly an ideal situation – whether or not it would represent even more of a pull to the centre than the current situation I could not say, though: the current system, with its obsession with a tiny number of 'swing' voters already has this effect pretty powerfully. I would argue that it also leads to the argument being carried out on entirely phony grounds: swing voters aren't renowned for their deep political analysis; on the contrary, I would suggest that someone who doesn't know who to vote for is more likely to be ill-informed than someone who does, and a system that pushes all parties to obsess on the opinions of a tiny group of ill-informed people is pretty much the worst of all possible worlds.
Of course, the effect having been in coalition with the tories will have on the lds is hard to factor in – these stats from past elections may have little bearing on the next election.