FTP

By definition, if you can sustain "a much higher power for a ride lasting over 60 minutes" then the FTP value you are looking at is wrong. FTP is defined as the maximum power you can sustain over a given period (typically 60 minutes) and so if you can sustain a higher power output, then that defines your FTP. 

I would guess your problem comes down to when and how your Garmin is measuring FTP. Ideally you would do a dedicated FTP test. But it looks like Garmin can be set up to auto-update based on any recorded activity (ref.). However, I note that the auto-update feature requires the new FTP to be higher than the previous one. If you've averaged a high power over >1 hour, I would expect your Garmin's FTP should update to the highest FTP based on the highest 20-minute window within that ride. That does require certain conditions to be met - so it's possible you didn't tick all the boxes on your more powerful ride?

However, none of that is relevant to the question: "Is FTP20*0.95 a good estimate of FTP60?" The answer to which is it's probably a reasonable estimate (see e.g. study).

I would also add that with most training metrics, being accurate in terms of the absolute value is rarely useful - it is far more important to be consistently measured so any trends or patterns can be identified.

Dr Coggan said that FTP and resulting power zones should be based on power generated in races etc out in the field, but in the absence of that the tests suffice so his preference was for actual one-hr power.

Assuming you are using the FTP for training, you'll soon know if its right when you try some sessions.

Personally, I've found the ramp tests a better guide than 20 min tests.

Yes. FTP is the maximum power you can average over 60 minutes. The maximal 20min x .95 is just a heuristic to try estimate the 60min FTP. It's not a terribly reliable heuristic either. If you have 60 min figures that are higher than the 20min x .95, then it is that 60min figure that counts.

I may have misunderstood your question but to be clear a 20 minute FTP test is an all out no holds barred - feel sick at the end - effort which is then multiplied by .95 to estimate your 1 hour FTP.

Not the same as just taking 20 minutes of an average ride.

This is just an estimate though and depending where your strengths lie you may be able to produce a significantly higher number at a full 1hr test.

FTP itself is of course just a guide number useful to gauge efforts for various training programmes.

Other measures are available including perceived effort and I recommend using a mixture.

Yes.