As mu_eff goes down (air gap goes up), the ideality of the geometry falls. That is, fringing and leakage are more and more significant.
So, you'll have higher A_L than predicted from straight (naive?) geometry, A_L will depend on the position and size of the winding, etc. Losses may also be increased, due to eddy currents induced by the fringing fields (so, especially near the gap, where "near" means within a couple widths; for a gap this wide, this certainly penetrates the bobbin, so will be important to the design).
Likewise, the flux density seems anomalously low, because much of it is bypassing the core and looping around (and within!) your winding instead.
This is one of the advantages of toroids, that the air gap is distributed; you also don't have anything else to calculate, just pick the right part for the job, no air gap to cut (removing a degree of freedom isn't exactly a net positive, it just makes things more straightforward
). The downside is they're much lossier than ferrite, for the most part; or if you choose the low-mu RF types, losses can be reasonable at low frequencies (the ~MHz used in high performance SMPS), but it'll be bulkier than an optimized ferrite design. The catch is, because of the fringing fields in the ferrite case, litz wire is probably required.
And you can indeed get quite high Q factors from litz + ferrite; I measured Q ~ 500 on a recent design (RM12 in N49 ferrite, 0.1mm strand litz, ~1mm air gap). Saturation was also extremely high (about twice expected).
Tim