Strava Run vs. Garmin Connect

What was the total distance of the run? That .1 mile difference may be huge, or within reasonable margins.

My first guess is that the Strava app has an "auto-pause" function kicking in...arriving at a similar total distance, but returning a faster average pace because it subtracts out the time spent waiting around at stoplights.

Just eliminate the guesswork...go to a track and run on the inside lane. That will tell you which device is the best.

If you have a Garmin, or any other device with actual GPS, and upload the activity to Strava the Strava data is built from the watch GPS data so its accurate. However, if you are using a mobil phone device (iPhone, Galaxy, etc) this uses its digital phone signal for triangluation GPS, including altitude.  This is not as accurate as actual GPS.   If you use a footpod in combo with a phone device it still is only as good as the app GPS.  If you use both the phone device with footpod and Strava app on your phone they are both not terribly accurate and I would guess that fluctuations occur.  Its how they interperate the digital gps data - I guess.  Even with actual GPS devices there is slight fluctuation between riders/runners on the same rides/runs.  Perhaps, maybe, I think so...

This is absolutely not true. Most modern smart phones have an actual GPS chip. They do use cell phone triangulation for some apps where accuracy is not a concern (ex: weather). For GPS logging apps or navigation, they use the cell phone tower triangulation to get a quick fix, then the actual GPS takes over and dials in a rather accurate fix. Here is an example of a run tracked by my Samsung Epic 4g using the endomondo app (later imported to Strava): http://app.strava.com/activities/7699176

It is an out and back and note how both lines track pretty darn close. Phone GPS is actually quite decent. BTW, there is absolutely no cell reception anywhere along that track.

Cell phones do tend to lose signal and so the resulting tracks hops to the next spot it regains a fix. They can also get a weird point several blocks away. It looks like you instantly teleport .25mile then instantly jump back on your actual path. These were two annoyances that led me to get a watch, but when it worked, the phone was great!

Possible, but I had plenty of weird issues with it in an armband. I know people get the same in a bike jersey. Both should be "in the open" enough for good operation.

My theory is that the phone has so much stuff going on, that random idling processes sometimes bog down the processor and that it just can't keep up for a couple moments. A watch is a much simpler, unitasking, device.

It all depends on the quality of the antenna in the device and the power levels it runs at.

Example: I just bought a camera with built in GPS.  The antenna is located on the top of it and is even clearly marked.  The instructions very clearly state to not ever cover that spot with your finger or you will lose GPS signal.  There is little doubt that a better antenna or diverting more power to the current one would correct that, but engineers make decisions to balance battery life vs performance.  For what it needs GPS for, it's good enough.

Additionally, I believe that cellphones and stuff also have issues with polling intervals (or that is perhaps app driven, not sure) where they may not update your location frequently enough for solid accuracy.  Lastly, there are differences in the accuracy of the various bits of hardware used.  No Garmin device out there is anywhere near as accurate as a GPS device made by Trimble, as an example.... but Garmin is consumer level stuff, and Trimble is hyper-accurate commercial/gov't grade stuff.

Just my $0.02

.1 miles over 3.1 miles is about 3.2%.  That's a fairly high discrepancy mathematically, but in the realm of a 5k that would not be that important to me.  Your pace dilemma may be more important if you are using pace as a measure of intensity of training while training.  Otherwise you can also just take the total time divided by the distance, and you can verify the distance using one or more maps (google maps for instance) plus route mapping software (TP, Garmin, etc.).  But there again, the accuracy of the map over that distance might not be much better - I don't know.

Calibrating over a longer distance may show more consistent results.  If it the discrepancy gets worse, you have a real problem with either or both methods.  Over an HM route I do weekly, the calculated Garmin (910xt) distance consistently varies by .1 miles (<1%), which can be attributed to twists and turns as well as the inherent accuracy of the GPS within the watch.