I know that ISO is not the sole determinant of IQ but it does give an idea of IQ, see this link, the S95 is better at higher ISO (by quite a bit and images are crisper too) so I'm leaning towards the S95 as IQ is the most important aspect to me.
If one compares the DxOMark data for the DMC-LX3 as compared to the S90 or the G11 at:
http://www.dxomark.com/index.php/en/Camera-Sensor/Compare-sensors
The (same, I believe) image-sensor that is in both Canons performs (roughly) equally in terms of Dynamic-Range and Signal/Noise Ratio with the DMC-LX3 image-sensor, but at a (DxOMark "raw"-referenced ISO Sensitivity level) that is (around) +0.75 "stop" greater than is the case with the DMC-LX3 sensor.
In the case of the S90/S95 (which have a minimum F-Number of F=2.0, as opposed to the minimum F-Number of F=2.8 in the G11/G12), this (at full wide-angle) translates to an improved "light efficiency" relative to the DMC-LX3.
Here's the catch. At focal lengths greater than full wide-angle, the minimum F-Number of the DMC-LX5 varies from F=2.0 to F=3.3. However, in the case of the S95, the minimum F-Number varies from F=2.0 to F=4.7. As more and more telephoto optical zoom is used, the higher minimum F-Number of the S95 will give up all of the "light efficiency" advantages of the S95 (due to the loss of incoming light), and more (a full "stop" of light loss).
So, it would be fair to say that the S95 likely outperforms (at full-wide angle), but that + 0.75 "stop"
advantage gradually disappears completely at higher focal lengths, and appears to even become a slight (-0.25 "stop")
disadvantage of the S95 relative to the LX5 at full telephoto zoom settings of both cameras ...
It's interesting that (despite the S95's apparent slight SNR advantage over most of the telephoto zoom-range), canon (nevertheless) chooses to employ (what is reputed by many, but not observed by me personally) more aggressive in-camera JPG Noise Reduction.
Depending on your tolerance for lessened fine-detail (as a consequence of more aggressive NR), the S95 images may well look a bit "cleaner" than the LX5.
From my perspective (owning a DMC-LX3), I think that
both the LX5 as well as the S95 may (where it comes to in-camera JPGs) smear image-detail more than is to my liking. Thankfully, I shoot in "raw" these days ...

... which does nothing to mitigate the above-described trade-offs.