Right now Starship and it Superheavy booster are both open design. Current plan is for 35 engines on 1st stage, 6 on 2st stage, 150ton to LEO ascend payload, dry mass 85t, return to Earth payload 50t.
All of this could change during rocket development, because of technological challenges, but also money which will have SpaceX available during development process.
They already change Starship design many times ( number of fins, TPS types, design of legs ) and they could keep changing it constantly after first test flights. At the end they can finish with much smaller, medium version of Starship with less engines, less payload to LEO, smaller gross mass. For example 20,30t to LEO Starship/Superheavy with just total of 15 Raptors, but maybe 20,30t to LEO payload, but still with 30-40 Raptors total. It depends, on what would be optimal for lowest possible cost per kilogram to LEO, GTO, which will be main goal.
Figure 2 million per flight is just < aspiring goal > and shouldn't be taken as fact.
For example ESA Arianespace want with expendable Ariane 6 ( payload 20t to LEO, 10t to GTO ) achieve same cost per kilogram as F9R, which is about 5000$ per kilogram to LEO ( more than twice to GTO ) and with Ariane next program, which could use potentially reusable LOX/Methane engine Prometheus, improve Ariane 6 cost per kilogram further by factor of two. But that of course is not yet given.
Same as is not given that Starship/Superheavy cost per kilogram to LEO or GTO will be better than F9R, since second stage reuse is much harder than 1st stage ( both technically and economically ), and no rocket with massive size of Starship/Superheavy ( it will have see level trust 3-5 times of Saturn 5 ) was ever build, let alone try to be reused.