The Theoretical RTP (the figure published for the game) is calculated on one full cycle of a slot. In other words, when every combination on the reels has come up once, your RTP should be as advertised. Of course, since the spins themselves are random, each individual combination doesn't come up once and not repeat itself until all the others come up, so actually getting results that are consistent with the TRTP is difficult.
As you play more and more, your personal RTP will gravitate towards the TRTP as the law of large numbers etc kicks in. Now, you will obviously have times where you are ahead and behind, sometimes to large degrees, like a pendulum (variance dictates how wide the pendulum swings), but as time goes on your personal results will vary less and less as your wins and losses become a smaller and smaller percentage of your overall play e.g. if you play a slot for the first time and win $1000 on your first spin, your RTP will be huge overall, but after playing it for a year, another $1000 win will have a much smaller effect on your overall position.
The idea of winning $100 on your first $5 bet and then losing the next $95 of bets is erroneous.....not that it can't happen, but its not how things work. If slots did work this way, they would be unplayable, because you would only be assured of losing everything you won in the spins immediately following every time. You may as well have a slot that pays back .95c each and every time you bet $1.....you could never ever get ahead, but its the only way to ensure your RTP is spot on.
Just consider TRTP in terms of making decisions about which games to play. The higher the TRTP the better, but you can still lose your shirt on a 99% slot, as a result of variance....which someone else can take over explaining.