Yup, I notice a difference of 0.5 to 1 mph between Cyclemeter and Strava, with Cyclemeter usually registering me as faster. My best guess -- experimenting while riding in a car or bus and watching the auto pause/resume -- is differences in threshold for auto pause/resume. Cyclemeter appears to go into pause more quickly when I slow down or stop, and doesn't kick in again until I'm above something like 5 mph.

Wahoo Fitness offers user-adjustable thresholds for auto pause/resume, but I'm not sure how effective it is. I don't recall whether Strava and Cyclemeter have adjustable thresholds.

Anyway, that's my best guess for the differences in speed.

Regarding elevations, as far as I can tell they're all using a common database for reference. I don't think they're actually measuring our climbs and descents on the fly. But when I review the traced paths later they're obviously inaccurate -- showing me to appear to be rambling around across the roads into ditches, across barbed wire fences, etc. So I'm guessing the variations in distance, elevation gains, etc., are due to tracking errors.

That's one reason why I've occasionally stopped mid-climb to measure a slope using my iPhone on top of my horizontal top tube -- one advantage to these older steelies with horizontal top tubes! The grades shown on Google maps and others appear to be pretty accurate.