Lord Tim
After gating, this problem pretty much goes away entirely. It's an extra step but it can save HOURS of work moving stuff around...!

Actually it only helps with early and ghost detected markers. I recorded the metronome and activated Audio Snap. Still late and wrong markers.

Like I said, even with a generic pulse waveform like the internal metronome you get the late markers.

I'm with the poster who mentioned melodyne. I find I use it more and more for drum timing issues. I'm considering getting the full studio version of melodyne 4 for doing my multitrack drum edits.
 
It seems to me the idea of identifying and working with "blobs" is a better idea than transients. A drum hit is really more of a blob than a transient as a hit is bigger than a single transient. Although the transients should identify the beginning of the blobs, hence the start of the hit.
 
The two really go together, seems like identifying the blob would be easier, once you find the blob you can go to the beginning of it to find the first transient. This two step identification of transients might be something a baker would read and trigger an idea of how to do it better in sonar.
 
gabo

Yea just using it on single tracks, although if all of them are aligned to the grid you can do them one at a time and do ok. 
 
I've found with my drums that I can use transients first on the multitrack parts and only align things on the 2 and 4 beats and on the 1 beats at the end of runs/transistions, aligning those to the grid. This brings everything pretty much in time with a few anomalies here and there.
 
Then I bring in each track to melodyne and use it to fine tune the alignments. Since everything is really close from the work described above, it tends to line things up well across the tracks so they match. With melodyne you can work on sections of the tracks so it's easier. I wish you could do that with audiosnap.
 
gabo

Sounds with a higher frequency and fast attack (like most metronome samples) are less likely to exhibit late transient detection because the peak amplitude occurs almost immediately. The late transient markers are more common on Kick drums, Bass guitar and other low-freq and/or slower attack sounds.
 
I suspect the reason the Bakers have not responded is that the change in behavior of the detection algorithm was deliberate to reduce the number of false positives and early placements which it definitely has done compared to the old algorithm.
 
The question is which issue is more troublesome. As i said before, I generally found it easier to clean up the relatively few mistakes made by the old algorithm than to try to fix all the late markers produced by the new one, and if your goal is to split at transients or quantize audio, late markers are more problematic than early ones. Early marker placement can usually be addressed by gating the material first, but late markers cause clicks, distortion and doubled attacks when splitting/stretching, and there's no pre-processing you can do to avoid it.