Actually, the difference between the so-called "20 meter" ham band and 22 meters is less than it might seem. The phone portion of the "20" meter band is really much closer to 21 meters wavelength, and so the dimensions of the antenna elements will only be off about 5%...only around 2% if the Yagi is cut for the CW end of "20".
That's enough to throw higher VSWR numbers than you'd want for transmitting, but as K3ZRT said, it won't blow anything up when receiving. It's not a major loss of efficiency either, so long as your receiver's input impedance is a true 50 ohms or greater. As for directionality, I wouldn't worry at all. The difference will be indistinguishable to the ear. Such a small error in element spacing will slightly affect the depth of nulls and gain of the forward lobe, but has very little effect impact on their orientation. At least, I've found that to be true with a 3-element Yagi; the more elements the antenna has, the more critical spacing could become, and the more bands it covers, the more chances for the traps to complicate its out-of-band impedance characteristics.
Ed makes good points about isolating the transmission line for control of common mode noise. And, reducing common mode coupling may also have beneficial effects on directional properties of more complex antennas as well.