The limitation isn’t the PCB, it’s the processor. The way IoTaWatt works is to sample the channels round robin, one cycle each. The ESP8266 algorithm requires a half cycle (8.3ms@60Hz) between samples to analyze the previous sample, do everything else, then setup for the next sample. So it ends up capable of a maximum 40 cycles samples/second, resulting in each channel being sampled about 2.5 times/second.
You could spread it a little thinner and add maybe eight more channels, total 22, but there are no unused output pins on the ESP8266 for an ADC chip select.
That said a lot of new users anticipate measuring way more circuits than is necessary or practical. For one thing, the 240V circuits use two panel slots but can be measured with one CT. In this era of LED lighting, many lighting circuits can be combined through one CT and the total power is still relatively insignificant. You will have a number of circuits that will be zero most of the time. If you don’t measure them directly, you can define an output that is the total of the unmeasured by subtracting the measured circuits from the panel main. For example I have a radon mitigation fan that draw 30 watts 24/7. I don’t measure it. It’s part of that “unmeasured” metric and contributes to it’s baseload, which is nearly always a flat line. When it spikes, it’s usually something identifiable like the garage door opener that operates a few times a day and uses a trivial amount of power in the big picture.
Look at some of the installations in this forum. Quite a few folks will tell you that they get a high percentage of their usage measured accurately with 14 channels. Also, don’t underestimate the cost of 40-60 CTs.