I am far from an expert on on this backup tool although I have been using it for about a month now on a new Linux Mint system. First of all, you need to understand that Deja Dup is simply a graphical front end for the tool that actually does the backup - duplicity. Duplicity works by scanning what you have defined to be backed up and searching for differences since the previous backup. The differences are then compressed (and optionally encrypted) and stored in the current backup directory. Thus, the whole process starts with an initial full backup and continues from there with incremental backups.
With that in mind, I'll try and address your questions.
1) As mentioned above, the first time you do a backup with duplicity, it backs up all files that you have defined. These files are all lumped together and the resulting backup file could become too large to easily work with. Thus, Duplicity splits it into multiple smaller files (around 50MB) that are individually easier to work with. Once the initial backup is complete, subsequent backups only need to track what has changed and can be significantly smaller. The downside to this approach is that if you want to restore from a specific date, the restore needs to start with the original file and then apply all changes since then until it gets to the target restore date. Thus, it's probably a good idea to periodically do a full backup so that a restore doesn't have to go back so far. I use duplicity directly on a Raspberry Pi to backup weather station data and the script I wrote to do it creates a new full backup at the start of every month. Thus, worst case, restoring a specific data set would only require retrieving the original and applying about 30 delta files. Whether Deja Dup occasionally restarts with a full backup, I don't know. It hasn't so far and I will be keeping my eye on it over the next month or so to see if it's a "keeper".
2) As I mentioned above, the backups are compressed so it's not surprising that they are smaller than the directories being backed up. Also, I just looked at my backups, which are stored on a NAS - the first backup consisted of over 1500 files on 03/28/2018 (all 52.4MB). Since then, it is typically a few backup files per day, although some of these are much larger (I saw one that was almost 900MB).
I should also point out that I arrived at your post while searching for a question of my own - how do I restore a single file or directory using Deja Dup? I basically want to go back and look at a couple of files that I deleted a couple of weeks ago, just to compare them to what I'm using now. I could find no way in the rather sparse Deja Dup app to do it. As near as I can tell, it wants to restore my entire backup (fortunately, it can be directed to restore someplace else other than the backup location. However, I have no desire to restore 500GB of data just to look at two files. I'm pretty sure I can do this directly on the command line with Duplicity, so if Deja Dup doesn't support this I will be extremely disappointed and will likely either switch to a different backup tool or write my own. I would also recommend this to you as well - try to restore something and see if you are happy with the results. If not, you can then start looking for a new backup tool before you find out that your current backup tool let you down.
One last comment - you can learn more with "man duplicity". You will find that each backup consists of three file types: a manifest file, signature file, and one or more backup files in compressed tar format and optionally encrypted with GnuPG. There are also a few good tutorials on duplicity that you can find with a web search.