There is clearly something going on with that amp that allows it to output 35 watts from a pair of EL84s when no other amp is even close to the figure. In fact, there are no other EL84 amps that will make more than half that power from a single pair of EL84s. Modjeski is a brilliant designer and perhaps his point applies to his unique topology used in this amp.
Yes, a genius designed it. Three things the man is known for are: quiet amps (quietest tube amps I have heard), deriving more power from tubes than others have, and electrostatic speaker design.
In all seriousness though Roger specifically states he wrote his own application for the EL-84 tube. The original concept for designing the amp was to power his Quad ESL 57 speakers (not many know this but Roger cut his teeth working for Harold Beveridge designing the ESL amp for those speakers). Clearly the typical EL-84 application wasn't going to cut it.
For those inclined, the 6 Moons review of the RM-10 MkII is a good read with interesting comments from Roger, including the one below, as well as theories on light loading and 2nd harmonic distortion. I'd also recommend reading the RM-10 MkII manual for additional insights from Roger.
http://www.6moons.com/audioreviews/musicreference/rm10.html
"Over the last 10 years, I have written several new applications to get higher power out of some common tubes that I thought could do more. If I were employed by Sylvania in the 1950s-70s, I would have done the same thing. There are an infinite number of applications that can be written for any given tube. They involve variations in B+ voltage, current, grid drive, screen voltage and load impedance. The data books list only a few to give amp designers some ideas. Sadly, that's as far as many designers go. David Manley was a stickler for applications believing that the only ones allowed were the ones in the data books and they must be adhered to the letter. Sadly, he took the ambitious 560-volt ultra-linear application from MOV as reliable and made some amplifiers that damaged a lot of tubes. Not even MOV could make tubes that consistently held up in that application. To this day, there are no tubes that hold up well at those voltages on the screen.
I research my applications very carefully with dissipation studies on all elements. And, although I have a replacement tube business, I do not want to waste valuable tubes (or your money) on poor applications. One application that caused a lot of comment in the Asylum was my getting 35 watts out of a pair of 6BQ5/EL84s [the amplifier under review - Ed]. They said it couldn't be done yet it is done and has proved reliable over the past 10 years in the field. The RM 245 uses a new application that yields 4 watts at a dissipation of less than 8 watts. There is no extra stress on the tube. Typical applications are 10 watts dissipation (the max for that tube). If you look at the following link, you can see the power vs distortion for 3 tubes at 25 mA. and the EML at elevated current where we take advantage of some of its extra dissipation. I have not yet written the maximum application for that specific tube but I expect to get 8-10 watts at low THD.
The above referenced graph of power output vs THD is at the load. You can add about 20% at the plate. The commonly used application from the RCA manual (RC-11 pg. 86, 1933 and repeated thereafter without modification) gives 2 watts at the plate with unspecified THD (shall we assume 5-10%). Two watts was more than enough power for the times and it seems little has been done by modern 45 amplifier designers to consider new applications.
Applications once written are forever left to rest and copied verbatim by every other maker. Remember, an application is often written to entice the designer (of a table radio at that time) into a circuit that fits his technical and economic needs. The low B+ (275V at 36 mA = 9.9 watts dissipation) of this application is attractive. Note they have 2 even lower power ones for those more economical who desired less power back then. This tube was designed to be a step up from the type 71 (max pwr. 0.79 watts) and a practical alternative to the dismally inefficient and hard-to-apply type 50. Note that the 50 was rapidly pushed aside for the much more popular 45. I have a Brunswick Panatrope that uses a 50. It's a monster that sold for the price of 1.5 Model T Fords in 1928. The 50 also had to be transformer-coupled when resistance coupling was coming on fast. My application is 310V at 25 mA =7.75 watts along with circuit and load differences.
I "light load" my RM-10 and by default have no choice but to use the 3 ohm tap on my Music Reference EM-7 monoblocks (my speakers are 8 ohm minimum with 12 ohm peaks). I'm not hearing anything negative like lack of bass, quite the contrary with the RM-10, I thought the bass improved. Having met the man a number of times when I lived in Santa Barbara I can vouch for his engineering knowledge. What he does to make his equipment sound so good is beyond me though. I just know it's relegated some really good equipment to the closet.