LIke `ping`, `tracert` doesn't really measure latency. It measures the time between a (UDP) TTL probe and an ICMP response. This can be close to the real latency but due to additional delays also larger by an unknown factor. ICMP generation has a low processing priority and can be delayed or even dropped depending on current load, previous messages, configuration, ...
Apparently, hop 2 takes its time sending the ICMP response - this increases the response time by at least 50 ms.
Hop 7 doesn't respond at all, the timeout is displayed as `*`. Hop 5 doesn't respond to all probes or one of them (or a response) got lost.
Real latency (which you only get a hint at with `tracert`) mostly increases when a link is congested and packets need to get queued. Also, elaborate encoding on a link or just the propagation speed on long links add significant delay.