So, I should clarify something: A visual approach is pretty simple from the controller's point of view; it's more work for the pilot. But, the pilot is always the one in command of the plane: If the pilot is unwilling to accept a particular approach, they just need to say so. It may lead to delays, but controllers will accommodate.
I originally said that the pilot had been assigned a visual approach to runway 28 Right. Looking at the charts for SFO[1], and the FlightAware trace[2], I have an alternative: The pilot may have been flying the LDA PRM runway 28 Right approach[3].
(You should open up [3] to look at while reading the rest of this comment.)
As has been mentioned, SFO is a bit weird: The two main runways are close enough together that it's not possible for you to have normal simultaneous approaches. For example, have a look at the diagram for Cincinnati/Northern Kentucky International (CVG)[7], another Class B field. Now, compare that to SFO[8]. At CVG, the parallel runways are far-enough apart that simultaneous approaches are possible, but with SFO, that's not possible. So, the FAA came up with the PRM (perimeter) approach.
Essentially, the plane approaching runway 28 Left is on an ILS approach. The plane on approach to Runway 28 Right is on an approach that is kindof like an ILS approach, but the plane isn't heading towards the runway, the plane is heading to a point to the right of the runway. Notice on [3], the point labelled "CFFKC". That's a fix (a point in lateral space), marking the point that you're being guided to. Note how the fix doesn't line up with the runway; that's why this approach is an "LDA (Localizer-type Directional Aid) with Glideslope" approach.
As per the detailed instructions (in [4], and point 4 in [6]), pilots on the LDA PRM runway 28 Right approach switch to a visual approach after the DARNE waypoint (3.4 nautical miles from the runway end). For most of the approach, they are monitoring a second frequency, which is only used if planes get too close together. But, once you pass DARNE, you no longer need to listen on that frequency. And although tower might have a radar view of surrounding traffic, it's not their primary tool (their eyes are).
If the Air Canada pilot was flying the LDA PRM RWY 28R approach, I think the final sidestep (from the LDA to the runway) may have been missed, or not completely executed, causing the pilot to line up with what he thought was the runway, but what was actually the taxiway.
Wow, that's a lot of detail and analysis. That does make a lot of sense. I know it will be a while for the FAA report to come out, but if it does turn out to be what you've stated, you can totally say you called it.
By the way, for anyone still interested a month or so from now, check out NASA's Aviation Safety Reporting System site [1]. This is a place where pilots, air traffic controllers, and other aviation personnel can go to anonymously report things that they were involved in. Reports are confidential [2] (they're anonymized before publication), and reports may not be used for enforcement purposes [3].
I expect that the pilots and controllers involved will have already submitted ASRS reports of the incident, and they'll eventually pop up on the site.
I think many of the readers here would be interested in virtual ATC, as many parts of the brain: Keeping many things in order, looking multiple steps ahead at what's going to happen (it takes time for a plane to turn), and communicating with your "customers".
The only downside is, you shouldn't sign in to control when you're on call 8-)
I originally said that the pilot had been assigned a visual approach to runway 28 Right. Looking at the charts for SFO[1], and the FlightAware trace[2], I have an alternative: The pilot may have been flying the LDA PRM runway 28 Right approach[3].
(You should open up [3] to look at while reading the rest of this comment.)
As has been mentioned, SFO is a bit weird: The two main runways are close enough together that it's not possible for you to have normal simultaneous approaches. For example, have a look at the diagram for Cincinnati/Northern Kentucky International (CVG)[7], another Class B field. Now, compare that to SFO[8]. At CVG, the parallel runways are far-enough apart that simultaneous approaches are possible, but with SFO, that's not possible. So, the FAA came up with the PRM (perimeter) approach.
Essentially, the plane approaching runway 28 Left is on an ILS approach. The plane on approach to Runway 28 Right is on an approach that is kindof like an ILS approach, but the plane isn't heading towards the runway, the plane is heading to a point to the right of the runway. Notice on [3], the point labelled "CFFKC". That's a fix (a point in lateral space), marking the point that you're being guided to. Note how the fix doesn't line up with the runway; that's why this approach is an "LDA (Localizer-type Directional Aid) with Glideslope" approach.
As per the detailed instructions (in [4], and point 4 in [6]), pilots on the LDA PRM runway 28 Right approach switch to a visual approach after the DARNE waypoint (3.4 nautical miles from the runway end). For most of the approach, they are monitoring a second frequency, which is only used if planes get too close together. But, once you pass DARNE, you no longer need to listen on that frequency. And although tower might have a radar view of surrounding traffic, it's not their primary tool (their eyes are).
If the Air Canada pilot was flying the LDA PRM RWY 28R approach, I think the final sidestep (from the LDA to the runway) may have been missed, or not completely executed, causing the pilot to line up with what he thought was the runway, but what was actually the taxiway.
[1]: http://www.airnav.com/airport/KSFO
[2]: https://flightaware.com/live/flight/ACA759/history/20170708/...
[3]: http://155.178.201.160/d-tpp/1707/00375LDAPRM28R.PDF
[4]: http://155.178.201.160/d-tpp/1707/00375PRMAAUP.PDF
[5]: http://155.178.201.160/d-tpp/1707/00375PRMAAUP_C.PDF
[6]: http://155.178.201.160/d-tpp/1707/00375PRMAAUP_C2.PDF
[7]: http://155.178.201.160/d-tpp/1707/00655AD.PDF
[8]: http://155.178.201.160/d-tpp/1707/00375AD.PDF