Checking my latest track on RealSpeed I noticed an unusual discrepancy between track point and Doppler of 5 knots on my last run. The sudden surge in the Doppler speed as I pulled up can no doubt be accounted for by the change in satellites and HDOP but what about the rest of the 10s? I am thinking as that part of the track is a short slingshot over mirror flat water with me hiked out almost horizontally to windward the gps is getting a reflected path from the satellites but this does not affect the trackpoint calculation. Any ideas?

So what's the yellow line? that has an inverted "V" underneath the end of you doppler peak.

Nº of satellites I think

I'd be willing to look at the data... but from the snippet, I wouldn't be confident of finding anything.

Hi Yoyo. I think you have found the dreaded Doppler shift!

There have been some well studied cases like this in the past. Ones form Chris Lockwood and Steve Thorpe com to mind. In Steve's case he has a second GPS to compare it with so it was easier to deal with.

The phenomenon is seen where there is a sudden step up of Doppler speed, usually for the duration of a run, where it suddenly reads 5 knots or so higher than the Trackpoint speeds, or in Steve's case, the other GPS.

We don't really know what causes it but we have some theories. My best guess is a jump of one satellite to another where the position of that satellite has not been updated or has wandered. It could be a error in the GPS device itself. In any case it is quite rare and as you saw it is pretty easy to spot if you look at the positional trackspeed difference, and very obvious if you have a second GPS to compare.

As a secondary obsevation, this is exactly why we say the Canmore is less reliable. If it has this error, there is no way of knowing as it does not record the trackpoints or sat info.

I have read that while the gps may track 7 to 12 satellites it uses many less for the calculations (I read 4..but can you trust what you read on the internet?) . In any case it may be that while the number of satellites recorded remain unchanged (7 in this case) maybe the satellites used for the calculations changed.
I guess the answer is, as you say, use a backup.