Wow, that went sideways fast...
Anyway, I took a look at one of your subs.
To get an accurate image scale, I plate solved it. For some reason, the ImageSolver script failed to solve it, so I used PhotometricColorCalibraton. To get it to work, I converted the image to the RGB color space, because that's a requirement of PCC (I could have debayered it). I used the Search Coordinates button and used M65 as the target object. Then I entered 560 as the focal length and 4.63 as the pixel size. The date is not hugely important, so I just change the year to 2019 and didn't worry about the month and date (it would have worked, even if I left the year at 2000).
I then applied the process to the image. I let it run until the solution was done and then aborted it. Here is a paste from the Process Console window that has the information I was looking for:
Loading image: w=4144 h=2822 n=1 Gray UInt16
7 image properties
48 FITS keyword(s) extracted.
* Loaded astrometric solution:
Reference matrix (world[ra,dec] = matrix * image[x,y]):
+4.44020578e-04 +1.67690806e-04 -1.15661383e+00
-1.67509829e-04 +4.44012659e-04 -2.79412742e-01
WCS transformation ...... linear
Projection .............. Gnomonic
Projection origin ....... [2071.989700 1410.976400]px -> [RA: 11 19 00.097 Dec: +13 17 17.55]
Resolution .............. 1.709 arcsec/px
Rotation ................ -159.324 deg
Transformation errors ... ex=9.55e-12 ey=2.5e-12 px
Observation date ........ 2019-04-11 08:56:59
Focal distance .......... 558.96 mm
Pixel size .............. 4.63 um
Field of view ........... 1d 58' 0.2" x 1d 20' 21.5"
Image center ............ RA: 11 19 00.099 Dec: +13 17 17.58
top-left ............. RA: 11 14 15.242 Dec: +13 00 21.97
top-right ............ RA: 11 21 47.945 Dec: +12 18 49.82
bottom-left .......... RA: 11 16 10.901 Dec: +14 15 38.41
bottom-right ......... RA: 11 23 45.613 Dec: +13 33 53.35
The important part of this for determining FWHM is the resolution of 1.709 arcsec/px. This is the number that you use as the Subframe Scale in SubFrameSelector.
Next, I ran the FWHMEccentricity script. I've pasted the output here. The important parts are Median FWHM of 4.498px and Median Eccentricity of 0.6592. With a little math, we can see that the FWHM in arc seconds is 4.498 * 1.709, or 7.7687 arc seconds. I then ran SubFrameSelector on this image and it evaluated it as 9.2334 arc seconds.
Note that this is not quite an apples to apples comparison. The FWHMEccentricity script computes the median, while SubFrameSelector computes a weighted mean. For the mean to be larger than the median, it means that you probably have some pretty large stars in the corners that are bringing up the mean.
In any case, as a final reality check, I created a preview of M66 and extracted the debayered luminance, and found the focus to be pretty soft. So I think that an assessment of FWHM between 7 and 9 arc seconds is probably correct.
Oh, and SubFrameSelector computed 0.6295, so pretty close to the FWHMEccentricity script.
If you want to explore the difference between the mean and median FWHM values, you could try debayering the original FITS, cropping it to a square that eliminates the corners and edges of the original, extract the luminance, and run both the script and SFS again. I would bet that the FWHM values are closer.