@Elizabeth Martinez great question, and exactly the right way to approach this — working from the link budget up rather than just trusting the headline range figure.
To answer the PHY question directly: yes, you're looking at two different operating modes. The 2Mbps spec is the peak data rate the module supports, but the 50km drone-to-drone figure is achieved at the coded PHY (125kbps S=8 coding scheme), where the sensitivity gain from the forward error correction is doing a lot of the heavy lifting. Those aren't in conflict — the module supports both, you just can't have 2Mbps *and* 50km at the same time. That's a fundamental BLE 5.0 tradeoff, not a marketing sleight of hand.
On the sensitivity figures — I don't want to throw out numbers and have you build a link budget on something I'm not 100% certain of for this specific front-end, because the external PA/LNA combination changes things meaningfully compared to a bare silicon figure from the Nordic or STM datasheet. The honest answer is you should get that directly from Moamen or the RFOXiA team through the Dev Hub or the contact on rfoxia.com, because the actual measured sensitivity with the amplifier chain in the loop is what matters for your calculation, not the chipset datasheet spec. That said, modules in this class with an external LNA typically land in the -103 to -108 dBm range at 125kbps coded PHY, which is where the extended range becomes achievable.
For your agricultural flat-terrain deployment, the good news is that's basically the best-case scenario for this module — the 50km figure is specifically the drone-to-drone case where both nodes are elevated and ground reflections are eliminated, but flat open terrain with clear LOS gets you much closer to that ceiling than any urban or obstructed environment would. If your nodes are ground-mounted you're realistically looking at the ground-to-ground figure which is 5km, so if you need longer spans you'd want to either elevate the nodes or think about a mesh hop architecture.
