In #2, her face might be overexposed like 1/3 of a stop or so...it's better to have it a bit over than a bit under on blondes. Anyway...I dunno...relying on TTL can be tricky because TTL flash often tries to "think for you", and it can think like a dumb machine, and can do things unwanted by you! The second photo appears more lighted by FLASH on her....the first appears like a mixture of flash and daylight, which has led to too cool of a color rendering on her, and weird sky tones...look like the exposure on #1 was bad in some way, and you rescued it, and in the process, got the sky coloration kind of, well, weirded out...
#1 her skin looks a bit greenish, and her eyes are dark, and the sky is messed up. Why? My *personal belief* is that this is what one can expect from spot metering in tricky situations...terrible results because the spot was not PERFECTLY placed, on the PERFECT part of the frame...in my experience, spot metering produces WILDLY INACCURATE exposures that shift huge amounts, based on very small camera angle changes, lens focal length changes, or minor mis-aiming of the spot...it is very common to have exposure readings that vary as much as four, five, or even six full EV values, with the aim of the camera moved only a few degrees!!! Spot metering is, when metering a backlighted sky tone, a really bad way to "peg" the sky value....a much better way is a broader-angle measurement from say, center-weighted metering.
As you know, there are TWO exposures in this scene. The ambient light (sky) and the flash exposure (softbox). ISO, shutter speed, lens f/stop, and flash-to-subject distance and flash output...five variables in all...SIX variables if one consider's the balance of flash-to-ambient light!!! ACK!
There's a LOT to discuss in this kind of scenario, using this method. Can't possibly fit it all into one post. One of the biggest issues though is the way EACH CAMERA MODEL meters TTL flash; the focus point in use can affect flash output on many cameras; some cameras are VERY sensitive to the actual AF point that is active, and will HEAVILY favor that AF point (Nikon D80 for example; many color-blind Canon bodies as well); using Matrix (Nikon) or "Evaluative" or Multi-Segment Evaluative metering (Canon, Sony,other brands) can drastically change the way the TTL flash exposure is computed, based on 1) reflectivity and 2) colors detected and 3) DISTANCE and 4) the computer chip's Matrix algorithms 5) Phase of the moon and 6)How few shots are remaining on the memory card and how crucial each remaining frame is. (okay those last 2, 5 and 6, are pure BS!)
The thing is this: TTL metering and flash control is often consistent, BUT, it can SHIFT on you! There are a multitude of weird variables in using TTL flash metering which can cause weird, unexpected shifts, like for example, the exposures might all look the same at 1/200, 1/160, 1/125, and then at 1/15, the "Matrix algorithm" will say, "HEY!!!!!! Let's cut the flash wayyyyyyy down, so it looks like dim-light, no-flash!" and then...you're hosed...