It depends a lot on the style of the song. For most of my own stuff, I'd be perfectly happy to use any old limiter that happened to be on hand. Seriously. When it's only chopping a few peaks here and there, and you don't set your brickwall right up to 0dB, it really doesn't matter which limiter you use. Heresy, I know. There's so much mystery - and strong opinions - surrounding limiters that's mostly unwarranted.
The exception is genres that rely on the limiter to smash the daylights out of a song. In that case, the limiter is touching the sound constantly, so even small differences in how it processes peaks can result in significant changes to the overall sound.
It's really kind of silly to expect a limiter to "preserve" peaks, when its very reason for existence is to beat peaks into submission. But there are some limiters that manage to mitigate the damage by analyzing the incoming material in order to calculate the optimum release time. Ozone is the best at this that I know of, though I'm told that Voxengo Elephant is its equal (if not as easy to use).
Pro-L, in addition to calculating program-dependent release times, also takes a novel approach to the
attack phase, separating the transient portion from the rest of the peak and processing the two parts with different algorithms. It also has a "Dynamic" mode that applies a transient enhancer prior to limiting, so that even though the peak still gets chopped, the waveform at least rises to that peak more steeply.
If you don't use the limiter as an effect but rather as a safety mechanism, then don't waste your money on a boutique limiter. Instead, use a cheap or free plugin such as Limiter No. 6 or TB Barricade, or SONAR's own limiter (just stay away from Boost11).
I you want to smash the daylights out of a track or sub-bus, the free
LoudMax limiter will do the job. The smash-everything crowd seems to like it on the master as well, but I wouldn't know about that.