I've seen discussion of this study before. While it's not surprising that the overheads with a larger team are somewhat greater, the discrepancy here is vast.
However, then you notice that the size of project being discussed here is only 100,000 equivalent source lines of code. Assuming that an "equivalent source line of code" is something like a real source line of code, if you're building projects on that scale with teams containing an average of 32 people then I'd say you're doing it wrong. Such projects are almost certainly simple enough for a single small team to complete inside a year, as indeed the study also found, and an insane proportion of the time spent by the members of your larger team is going to go on management, communication and integration overheads.
If you started looking at projects large enough that a single small team could not complete them within a useful timescale, perhaps an order of magnitude bigger or more, and you still found that teams averaging 32 members took almost as long to complete the work as teams averaging 4 members, that would be interesting. But that's not what this study seems to have considered.
If you start looking at projects significantly larger than that I'd argue that you should be leveraging service-oriented architecture and breaking the complexity up into manageable chunks that can be completed by smaller teams.
However, then you notice that the size of project being discussed here is only 100,000 equivalent source lines of code. Assuming that an "equivalent source line of code" is something like a real source line of code, if you're building projects on that scale with teams containing an average of 32 people then I'd say you're doing it wrong. Such projects are almost certainly simple enough for a single small team to complete inside a year, as indeed the study also found, and an insane proportion of the time spent by the members of your larger team is going to go on management, communication and integration overheads.
If you started looking at projects large enough that a single small team could not complete them within a useful timescale, perhaps an order of magnitude bigger or more, and you still found that teams averaging 32 members took almost as long to complete the work as teams averaging 4 members, that would be interesting. But that's not what this study seems to have considered.