Certainly $b\pm a$ have no particularly small factors. It would be interesting to know what results you have for moderate size $x.$ 

Here is some very weak evidence that suggests that, while $\mathcal{N}_{prime}(x)$ might be fairly large compared to what one could naively expect, probably $\mathcal{N}_{prime}(x) \lt \mathcal{N}_{mixed}(x):$ 

The product of the first 10 primes is $6469693230$ (with square root about $80434$.) This allows $512$ choices for $b \gt a.$ Of these, $106$ have both $b-a$ and $b+a$ prime. Here are the $a$ values:

3, 17, 57, 145, 209, 406, 418, 442, 462, 546, 609, 667, 759, 770, 782, 805, 874, 1015, 1131, 1254, 1330, 1463, 1870, 2233, 2346, 2431, 2717, 2730, 3135, 3230, 3289, 3315, 3451, 3570, 3705, 3770, 4147, 4370, 4485, 4830, 4862, 5278, 5423, 5474, 5681, 5865, 6270, 6279, 6555, 6670, 6783, 7293, 7854, 7917, 10010, 10166, 11339, 13398, 13585, 14007, 14326, 15249, 15470, 15834, 16302, 16422, 17342, 18734, 19227, 19285, 20010, 20930, 23023, 24310, 28014, 28101, 29393, 33649, 34034, 35581, 37145, 37961, 38019, 39585, 40755, 41055, 42427, 43355, 46410, 51870, 54230, 56202, 57057, 58058, 62491, 62985, 64090, 66990, 67298, 67830, 68034, 70035, 74613, 75922, 79373, 79534

**ALSO** In none of the cases are both $b-a$ and $b+a$ prime powers (except the ones with both being actual primes.)

---
It is probably no help to you, but I am reminded of the quirky paper [Primes at a Glance][1]. In case $ab$ is the product of all primes up to $p$ and $1 \lt b-a \lt p^2$ (or I guess even the square of the next prime) then $b-a$ is clearly prime. That paper shows that such an occurrence is rare and discusses related matters. 

  [1]: http://www.ams.org/journals/mcom/1987-48-177/S0025-5718-1987-0866108-3/S0025-5718-1987-0866108-3.pdf