Is audiosnap unreliable for everyone or I'm alone?

Usually when transients are detected early it's because there's some noise preceding the main attack. It's more of a problem when the space between the transients is dead silent, because then even deviation of a couple bits is detected as the start of the transient. It's less of a problem with a live signal having a significant noise floor that swamps those little deviations.
 
I suggest you use Process > Remove silence (without splitting) to gate out everything below about -48dB (the defualt IIRC) on the attack, and below -60dB on the release with a Hold value of about 50ms. That way the first thing the detection algorithm sees will be a sample at a level that can only be part of an actual transient attack.

karl rose also has a great audiosnap video on youtube
http://www.youtube.com/watch?v=weIZy_ibH5c
 
http://www.youtube.com/watch?v=7VR7WkHywTo
 
cakewalk:
 
http://www.youtube.com/watch?v=fwaXi540fZQ
 
 

Audiosnap kind of is what it is and the stuff you are describing... well you're kind of stuck having to find workarounds. What I would suggest is increase the Audiosnap Pallette Transient Detection Threshold up high enough so that even if those unwanted transients are being included at least the ones you actually want are being tagged with a marker as well. Then (as laborious as this is) go through and disable the ones you don't want. That way at least you are working with detected transients in their natural Sonar detected spot. Also you are getting the markers added across multiple tracks (if you are working with multi track drums). Then you can use the feature (that I totally forget the name of but it is detailed in the Drum Production webinar available through the Cake TV links) that groups transients across tracks so you can edit them all at once.
 
One thing to note though is with multi track drums you kind of WANT the transients that occur from bleed to be editable in Transients mode. If you leave those transients audible but do not group them and move them in conjunction with the close mic'd track you risk causing phasing and weird crap. If you want to avoid this and do not want the bleed to be audible in your mix and/or don't want to have to group those bleed transients across all tracks apply a gate to each track then bounce them to remove the bleed. After eliminating the bleed (like on your close snare, kick, toms, etc tracks) then you should get a cleaner set of transients and you do not have to worry about screwing things up in regards to phasing on your close mics. You'll still want to be moving your overhead and room transients along with the close mic tracks transients though.
 
The other way is obviously to keep a lower threshold and add markers as needed but doing this manual again can be weird and it is indeed very tedious.
 
Basically this type of work is plain old tedious and Audiosnap is kind of a bugger as it is. My problems with it in the past were performance issues though. It just choked all the time and did not work correctly. Supposed in X3 it is working more like it should but I have yet to test it out.
 
Other programs handle transient manipulation/time correction much better than audiosnap does and I hope that Cake comes out with Audiosnap 3.0 in the next version or completely dumps the whole AS concept in favor of how things are done in other programs. It just is not one of the finer points of the program at this point in time. Perhaps it was revolutionary when it was first introduced but I wasn't around back then.
 
Some hope though, when X3 was first released and I was still pissing and moaning about X2 I mentioned in a thread to one of the Bakers that we need an Audiosnap improvement. He seemed receptive to the idea so maybe, perhaps, possibly, hopefully we'll get it. I am pretty darned happy with X3 at the moment so X4 will likely be a very tough sell for me. If they got AS up to par with the other DAWs though that would probably be enough to make me upgrade all on its own.

I've given up on Audiosnap since it seems to always take so much time to even get basic transient detection right. I switch to Ableton when necessary. Now with Melodyne it's even more trivial.

I liked the noise gating trick though! May give that a try next time.

Sanderxpander
I've given up on Audiosnap since it seems to always take so much time to even get basic transient detection right. I switch to Ableton when necessary. Now with Melodyne it's even more trivial.

You have to be careful when using Melodyne to Quantise Audio in Sonar, whether you are using the Quantise Macro, or manually alighting notes on the Melodyne Grid. There is a bug that is often not very noticeable, but can be major if you're not careful.