As Uav_mapper stated, the online services are not nearly as good as being able to process in house and have control over your data. If your just getting started they might help with getting your feet wet but if you are serious about learning how the process actually works then an in house solution is your only option. Photoscan has a big learning curve but is worth it in the end, pix4d also has a desktop version that is a little more user friendly but in my opinion photoscan came out on top so I went with it. If you don't have a very capable workstation to run the program on you'll need to consider that cost as well since either program is VERY CPU, RAM and GPU intensive. I ended up building a custom workstation since my CAD workstation didn't cut it. Photoscan offers a 30 day trial and I think pix4d does something similar(at least they did when I got into this) so start with one, work with it for the month and then try the next to see which you prefer.
I know you stated that you already captured some data sets but one thing to think about is the quality of the images you have, are any over/under exposed, what time of day did you collect the images, are there lots of shadows, did you achieve sufficient overlap, did you maintain a constant elevation AGL during the flight? Lots of the flight planning apps are customizable and that could be a good or bad thing...drone deploys app will let you set overlap down way under the recommended minimum overlap for photogrammetry (60side/80front) they also give you the option of "automatic camera settings" which can really cause a disaster of inconsistent imagery. Most apps dont even offer the ability to follow the terrain and achieve a constant AGL but if you have terrain with a large elevation change from one ond to the other you GSD could be what you want at one end and way over on the other... or on the other hand it could be the opposite that you have a smaller GSD and insufficient overlap at the other end depending on if you launch from the high or low end of the project. There are a lot factors to consider when collecting imagery and in the end if you don't think it all through you will have inconsistent results, especially when trying to achieve an accurate 3d model and not just a nice looking orthomosaic. GCP placement is another factor once you get there that depends largely on what type of terrain you're mapping. The best imagery will be from around mid day on an overcast day so there is even soft lighting and no harsh shadows.
As previously stated once you get past all that the next thing is to find a surveyor or some other way to get GCPs into your data, until you have control there is no way to check the accuracy of what you're producing. Teaming up with a surveyor would be the best option and probably the hardest unless you know one (or are one like me) but it could be done if you can prove to them that it would help them. Other options are Aeropoints(at 6
$6k for a set), or buying and learning how to operate your own RTK set up which is also a pretty hefty learning curve.