Looks like its time to fight math with math!
The number of tiles observed by an object with a sensor range of R is equal to 3R^2 + 3R (+1 if you include the central tile, i.e. the tile containing your ship, station, or colony), not pi*R^2; we're working with hexagonal grids, not circles. As an approximation, it's not terrible to approximate via circle, but the formula isn't that difficult to work out and it's not that difficult to use. This gives an area of 3997 covered by a ship with a sensor range of 36 and 1141 tiles covered by a station with a sensor range of 19 (these include the tile containing the ship and sensor). Furthermore, you are computing area per manufacturing point, not sensor "length" per manufacturing point, when you divide the area covered by the manufacturing cost.
Beyond that, I would question whether or not you actually care about how much area is observed. The primary thing that matters about sensor range is how much advance warning you can get, and for advance warning it is mostly the sensor range that matters (granted, because this is a turn-based game, you more or less need to be able to observe the tile in which a ship ends its turn in order to detect its approach, but this is still more a question of having something close enough to the tile to observe it than a matter of how many tiles are observed; two million tiles observed that aren't in the right place are worth nothing, except in that they provide knowledge of where the enemy is not). It is sensor range, not total observed area, which determines the minimum number of moves required to be able to attack a target without the target having any warning of the attack. It is sensor range, not total observed area, which determines how many turns of warning you have before an attacking fleet hits the observer (i.e. how much time you have to respond to a threat). It is sensor range, not total observed area, which determines how many observers I require to provide a target point with a minimum effective sensor radius*. It is sensor range, not total observed area, which determines how wide the outer edge of the sensor bubble is and thus how likely it is that a ship will pass into my observer's sensor bubble. I could not care less about having up to date information on every unoccupied tile within 5, 10, 19, or 10,000 tiles of every game object I own. I care about whether or not I detect an oncoming hostile fleet in time to react to it effectively, and for that it is sensor range that matters, not the total observed area. Even for exploration it is not the total observed area I care about so much as the length of the leading edge of the sensor bubble (which, under the system used in GCIII, is 2R + 1 tiles), as the maximum number of tiles I can reveal per turn while exploring is equal to the product of the number of move actions on my exploration ship and the length of the leading edge of its sensor bubble.
The later station sensors are unquestionably dirt cheap in terms of sensor range per manufacturing point; your counterpoint of "but manufacturing points per observed area is similar at best" is immaterial because observed area simply isn't something that you should care about. Total area observed is mostly immaterial; sensor range determines how many observers you need to monitor an area, how much advance warning you can get, how close you need to be to another empire's worlds or area of influence to monitor what it does within the region that the computer doesn't want you to send ships through or build stations or settle planets within.
I will also add that you did not compute the sensor range of the sensor ship correctly. Cargo hulls (and, for that matter, all other hulls) have a base sensor range of 2, to which the sensor range bonus is added. Therefore, the sensor ship described has a sensor range of 38, not 36. This is nevertheless immaterial since your assertion that it is the total area observed rather than the detection range which matters is nonsensical. The total area observed has very little to do with how much warning I have of an incoming attack and thus how much time I have to respond to it; since it is how much time I have to respond and over how wide a front I have at least this level of early warning that I care most about out of all the things that sensor coverage gives me, and since even for those situations where I care about how much area is observed (i.e. observation of the activity of foreign empires, for which what matters most is how close my observer needs to be to the targets of interest rather than how much total area my observer can cover, especially since if I require more than one observer to perform my surveillance it is the sensor range which most directly determines the number of observers needed), it is sensor range, not total area observed, that I care about.
*If you wish to make it so that a central tile has an effective sensor range of R with R being uniform in all directions, you will need your observers to be spaced around the central tile in such a way that at least two thirds of the tiles observed are observed by more than one observer, and of the no more than one third of the total tiles observed which are observed by only a single observer, only half will be on the far side of the observers when looking from the central tile, except for those tiles covered by an observer at a vertex of the ring. A nonuniform effective sensor range around a central point has the advantage of offering greater coverage for the same investment, but does not increase the minimum detection range relative to the central tile (it does increase the average detection range, but you're still looking at a function of the sensor range of the observers, not of the total area observed).
And lastly, sensor starbases often go obsolete. As your empire expands, you often no longer needs sensor within your borders as much. Meanwhile, your sensor ship can be moved to your latest border, or wherever you need it. It always has maximal efficiency.
This is trivially false. A sensor platform, be it a ship, colony, or station, is only at maximum efficiency if every tile it observes is uniquely observed by the sensor platform and observing the tiles provides valuable information; otherwise, its ability to observe those tiles is redundant (in the case of tiles observed by multiple observers) or unnecessary (in the case of tiles where up-to-date information is not of any significant value), and its sensor coverage is therefore not being used at maximum efficiency. Furthermore, while a sensor ship can be relocated so as to provide coverage where it is needed, it cannot be relocated instantaneously, and it will not always be able to move to its new location by way of a path which allows its sensors to continually provide unique information. Even if it can be moved in such a manner that its sensors will provide at least some unique information on every turn that it takes to move the ship, doing so will not always provide information which is valuable. I have no need of knowing about the present state of tiles 200 tiles from the nearest opponent; it is quite unlikely that there is anything there of any interest to me, so even if moving my sensor ship in such a way that it can observe those tiles while relocating is possible, it is not worthwhile unless doing so does not divert my sensor ship from a reasonably efficient path to its new location.
Even taking 'maximal efficiency' to mean hitting some efficiency target less than maximum efficiency, you still cannot guarantee that a sensor ship will always be operating at 'maximal' efficiency. There are always tiles where turn-accurate state information is of little value, and it's not unlikely that a sensor ship in the process of being relocated will be observing many of these tiles. It is further possible that there will be constraints preventing you from using a sensor ship at something approaching 'maximal' efficiency; perhaps you are disinterested in offending the Iconians by parking a ship in their area of influence despite the optimal placement for your sensor ship being some random point in deep space which for some reason is considered to have a predominantly Iconian cultural identity, or perhaps you're in the process of relocating a few ships and so several of the ships are not going to be providing information on tiles whose state was unknown to you but for those ships, or perhaps you've secured an exploration treaty and alliance with a faction which renders your sensor ships in a certain region entirely unnecessary for the foreseeable future and thus in a position such that they will not be providing valuable information for some time.