Hello again guys,
I had a busy week and didn't find time to share my findings.
Anyway, I waited without feedback much more than actually trying to do something
The solution for my problem was really just doing everything again and again, then one time it worked (the basic positioning, at least) as the TK-102. I didn't even made tests with the tracker on a vehicle after that.
I just stopped using the Coban and am now testing the original Xexun TK-103-2 (as TK-102 too).
Along these weeks, I received many responses from other people in other forums, and most of them did exactly this, stopped trying to get the Coban work and jumped to other test tracker. We are now testing the TK-103-2... at least it worked fine right on the second try, as a TK-102.
Anyway, I can bet that this Coban GPS-103A I have sends the speed in KNOTS, too. I sniffed some data using the windows server that comes with the tracker and wireshark, before trying Wialon, and the difference was clear in high speeds. The windows server reported the correct velocity in km/h.
There is no setup on the tracker to change this to km/h, at least not in the manual I have.
It looks like that when sending the info via GPRS it sends the raw data it gets from the chip -- the speed of the Coban on my bench was 0.00 only half the time, the other half it was 0.05, 0.11, 0.03, etc... When sending by SMS, it converted and rounds this down to 0.0 km/h, so it is pre-processed when sending by SMS.
The Xexun always gives me 0.0 km/h from both SMS and GPRS when stopped, making me think that it always preprocess the NMEA info that the GPS module spits out.