So maybe this is naive of me, but it seems like you could scan at 2 or 3 angles and, knowing the physical geometry of the lens system and book, determine which fragments were on the front and back of the paper. It would likely be fiddly, but worth an experiment!
X-ray lenses don't really exist, although you can produce focused beams (using either reflection or simply collimation). So, these methods typically involve scanning a sample and measuring some fluorescence line (which gets emitted in roughly all directions).
That site doesn't mention the spot size, but its likely much larger than the page thickness, so the iron signal from a tilted page probably wouldn't differ in a measurable way from having the page perpendicular to the beam. Keep in mind that they require ~24 hours to get even the images shown. Instead it sounds like they rely on compositional differences in the ink from the two sides and look at fluorescence from an element that is only present on one side.
It might be possible to get the other side by taking a difference between the Fe and Ca signals (after scaling appropriately).
Focusing x-rays is not impossible but very difficult. If you look at x-ray telescopes such as Chandra or XMM-Newton you'll notice they have extremely long, very high aspect ratio mirrors. This is because x-rays generally only reflect at grazing angles. At high enough energies you can use particle physics equipment because gamma rays will produce electron-positron pairs on impact (allowing you to then track their trajectories and work backwards to determine the direction of the gamma rays). Such equipment is enormously expensive and also very large so it's not particularly practical.
