That seems extreme. By using big groups you will lose a higher percentage of turns just moving to yet to be roaded tiles. Groups larger than 2 will usually be a waste.
Exactly. I usually try to build roads by just one single worker, and once the road is finished, I move additional nearby workers on that tile (with no loss of time) to help with the irrigation/mine. This seems to be the most efficient use of available worker-turns.
Let's look at an example: I have 3 workers and want to mine 3 grasslands.
a) Single worker approach:
Turn 1: each worker moves on a separate tile.
Turn 2-4: they build 3 roads
Turn 5: all 3 of them move onto one of these tiles and (in the same turn) start building a mine
Turn 7: the first mine is finished and all three of them start the mine on the next tile
Turn 9: the second mine is finished and all three of them start the mine on the next tile
Turn 11: the third mine is finished, the workers are free to seek new employment.
b) Stacked worker approach:
Turn 1: worker crew moves to first tile
Turn 2: they build a road
Turn 3: they start building a mine
Turn 5: the first mine is finished and they move to the next tile
Turn 6: they build a road
Turn 7: they start building a mine
Turn 9: the second mine is finished and they move to the next tile
Turn 10: they build a road
Turn 11: they start building a mine
Turn 13: the third mine is finished, the workers are free to seek new employment.
So we can see, it takes 2 full turns longer that way. A total of 6 worker turns are "wasted" (two in turn 1, two in turn 5 and two in turn 9, when three - instead of just one - workers move onto an un-roaded tile).
Of course there are exceptions. For example, if happiness is a big problem, I might move 3 workers to an un-roaded gem mountain in order to connect the gems in 4 turns instead of 10 turns. Or if food is scarce for a certain town, a banana jungle tile may also be worth sending a crew of 4 workers onto it to get the jungle chopped quickly.
Let me also add a comment on whether or not to upgrade units. I usually do it in two situations:
1. When I'm going for a quick military win. In this case you want to use both, shields as well as commerce, for creating a big army, and upgrades are simply the most efficient way to use money for military: you pay 3g per shield whereas with rushing you pay 4g per shield. So when you have reached the tech for the weapon you want to use (let's say Chivalry/Knights or Military Tradition/Cavalry, depending on map size and difficulty level), you stop research and produce a monster army as follows:
a) Turn 1: pillage your iron (or saltpeter) resource. Set all cities to produce horsemen. Restore the road on the resource.
b) For the next three turns (or how long it takes to produce the horsemen) just rake in the cash. (You need marketplaces and as many luxuries as possible...)
c) When the horsemen finish, you go into the city popup, right-click on the horseman and upgrade to Knight/Cavalry. They will already be "ready to go" in the same turn they were produced.
d) Go back to a)...
2. When I'm going for a research victory (UN or Space Race). Here I need only a minimum military to prevent my neighbors from getting any stupid ideas. Of course my cities need all their shields to build urgently needed improvements (libs, markets, courthouses, aqueducts, universities, Copernicus, Newton, Bach, etc.) The less shields invested into military units, the better. So I build an initial army and keep it up-to-date with upgrades.