You cannot say that the statistic prove it wrong as it only consist of 1 sample (your account). If you want to prove it ering eith statistic, you have to randomly pick at least 30 accounts as the sample.
Now about your data, you do not get a single HotM in 350 pulls. That is entirely possible. In fact I also experience it before finally acquiring Devana.
Lets do the math, the probability of not getting HotM in 350 pulls is:
(100%-1.3%)^350 = 0.987^350 = 0.01026 = 1.026%
Do you know the chance to get Mysterzo in Taverns of Legend? Merely 0.1%. That means not getting any HotM in 350 pulls is 10 times more possible than acquiring Myztero in a pull.
Statistic should stabilize but the that is not like what you think. It is not based on time but based on number of data instead. If you want the deviation to be less than 0.5% (fall in between 0.8%-1.8%), you might need thousands of data.
If your merge your data with all players data, it would stabilize into very close to 1.3%. This is because just like there are unlucky people, there are also lucky people, they counter each other and therefore stabilizing it.
As the probability of not getting any HotM in 350 summons is 1.026%, I will give an example with 7 coin throw which have 0.781% of not getting any head. Imagine if there are 128 people asked to throw a coin 7 times (all use the same exact coin), based on probability/combination theory, the expected outcome is:
1 people get 7 head
7 people get 6 head + 1 tail
21 people get 5 head + 2 tail
35 people get 4 head + 3 tail
35 people get 3 head + 4 tail
21 people get 2 head + 5 tail
7 people get 1 head + 6 tail
1 people get 7 tail
The person who get all 7 tail might wonder why it is always head, even claiming that the coin is rigged. However, when all data are combined, the number of head and tail are even, proving that the coin is not rigged, afterall it is the same coin.
It doesn’t matter whether the 7 throws is conducted in the same minute or spaced once a year. Again, stabilization is not about timespan/length, it is about number of data.