"But when I read online, many suggest that convergence can be determined by the number of packets lost in the transmission."
Well, much depends, I guess, what you mean by convergence. Further, packet loss, when it's impacted by a convergence event, would be a very gross or inexact way of measuring convergence time.
To me, a converged network is one where all the routers have an accurate representation of the actual current state of the network's topology. If there's any network topology change (which might be logical or physical) the network will need to re-converge to once again have an accurate representation of the actual current state of the network's topology.
Regarding using drops as a convergence metric, consider two routers with dual routed links between them, if running OSPF or EIGRP, I can have both links with same or different costs, but I change one of those links metric, I might shift traffic around on one of the links. Will doing this, cause any packet drops? It should not. Is this a convergence event? Well, the router you change will, with OSPF or EIGRP, "notify" its routing peers of the change, so they can update their representation of the topology.
Now instead of one router with two links, I have two separate routers, downstream of another router. Each of those routers has a link that will reach the same destination. Again, changing a cost metric, traffic may shift its actual path through the network. In this situation, unlike the prior, packet loss may occur, although no links are down. This because, unlike before, the upstream router, and even the other router with a link to destination, will re-converge so as to redirect that destination's packets to the "new" better path.
Interestingly, in the last example, whether there's packet loss or not, much depends on whether the topology change is "better" path now available, or you've lost the "best" path. (The latter, may cause routing loops while network re-converges.)
The only "true" way, to measure convergence time, would be somehow log the time of some topology event, and then log the times all the routers have correctly processed event. (Understand, depending on the topology change, correct operations might not require all routers to have converged, but the network isn't fully converged until all do.)
How to get the times, from each router, might require debugging options and/or advanced logging options.
Lastly, whatever time PT takes, likely would not represent actual times using "real" equipment.
Also, understand, actual convergence times also can much depend on how various OSPF and EIGRP Cisco timers are set, and what the events are. (For example, if there's a "flapping" event, Cisco has [increasing long] back-off timers, to delay notifying routers of the event.)
