In the early 90's I worked for a company where long life (5 yrs at room temp, 2 years at -20C) was manditory from CR2430 coin cells in a high pulse\low average current regime. We developed a custom test which emulated the pused operation and so could predict battery lifetime. There were large (2:1 - 3:1) variations between manufacturers for the "same" battery. The discharge curves were like fingerprints; it was clear which "brands" were mearly OEM versions from other manufacturers.
The TI link is:
Using a capacitor in parallel is useful. However, remember that you have different states in Ble. The duty cycle is not the same in all states, which means that you have to make sure that you capacitor is properly loaded in cases several TX/RX activities are done at short intervals. So a smaller TX/RX current is also an advantage there.
Coin Cells and Peak Current Draw
This is worth checking out. I can't post the link directly but it was also an EETimes featured article --- a Google search ought to be productive. Note that it discusses both capacitor optimization and differences among battery manufacturers.
These are results from one specific brand of batteries - it is worth remembering that it can change dramatically for different brands. We ran some tests and found that more complex battery packs with internal temperature sensors were even *faking* safety-critical data - the manufacturer had saved money by simply removing the temperature sensor and faking the results!
The graphs (and photos of the internal teardown) are here: http://bit.ly/tCTPVC
A reasonable-sized capacitor would help for very short pulses, but as the pulse length increases (at constant average current) the needed capacitor will eventually be too big. Well, one has to do the experiment to find where that point is.
You're not missing anything! It's astonishing that the authors don't even mention capacitors in parallel to improve performance, which is a time-honored strategy. I speak from personal experience, having determined optimal strategies for coil cell use in certain wristwatches, ones that use electromechanical displays.
I am probably missing something here, but if the issue here is pulse current drain on the battery due to intermittent high current use, then surely most of the issues go away if there is a capacitor in parallel with the battery in some way? This way the pulse current comes from the capacitor (which has a much lower IR) and the battery has a long time to top the capacitor back up before the next pulse.