From what I could find out, it looks like they're unusual cases produced by the obstacle clearance calculations. There's a lengthy thread on PoA about it and the main points I saw are:
- LNAV/VNAV and LPV approaches are designed differently (the specifications are in FAA Order 8260.58)
- The Obstacle Clearance Surface (OCS; an imaginary zone that must be free of obstacles) is shaped differently for each approach type
- The OCS isn't at a fixed angle: it depends on the designer's choice of glideslope angle (e.g. for terrain clearance) and - in the case of LNAV/VNAV - on the prevailing temperatures at the airport
- The two approach types have decision points at different distances from the runway threshold, which also affects the glideslope requirements
- The resulting OCS may be at a higher angle for LNAV/VNAV than for LPV
The FAA Order has a lot of formulae to work it all out, but apparently the result is that if the calculations come out in a certain way and there's an obstacle close to the runway, it can penetrate the LPV OCS but not the LNAV/VNAV OCS. That means the LPV DA then has to be made higher.
In the PoA thread, every example of an approach with LPV minimums above LNAV/VNAV has an obstacle close to the runway. But the consensus was that the LPV minimums being higher than LNAV/VNAV is still very unusual overall.
Finally, although this isn't exactly the same scenario, an old FAA article on GPS approaches mentioned that the minimums aren't always what you would expect:
While [varying the obstacle clearance for an LNAV/VNAV approach] occasionally results in minima higher than the LNAV minima, the added safety benefit of a stabilized descent outweighs the difference in minimums
(That all refers to altitude, of course. The other part of the minimums is the visibility, but that should increase anyway as the DA and/or category increases. A high, fast aircraft needs to be able to see the runway from further away in order to descend safely.)
I saw an excellent answer here: http://www.askacfi.com/3854/lnav-lower-than-lnavvnav.htm (when you look at the page look for 'LPV') and as it looks to me it has to do with the width of the area evaluated for obstacles that is taken into consideration.