* **Option A** is a relatively uncommon metaphor for representing these input constraints and will *increase cognitive load* of users trying to parse meaning
* **Option B** is probably more easily parsed, but *takes more clicks to physically accomplish*

In weighting them, I would distinguish between [Usability vs. Learnability][3]

* If you have *repeated users*, the new input format can be learned and quickly processed in the future by returning users

* If you have a *kiosk like environment*, then I would aim for the highest degree of discoverability between the two options, even at the expense of additional clicks.

----

Also, to iterate over **Option A**, as [IronBasset pointed out][1], adding a **visual separator between the radio selection and checkboxes** should add clarity and separation between those two distinct option sets.  

[![Option A - Improved][2]][2]


No matter the case, additional interaction logic should be added so that the appropriate state is synchronized and enforced programmatically, without users having to regress and undo a particular action.  If the top radio button is ever selected, it should deselect any checkboxes as those option sets are mutually exclusive.

**Further Reading**:

 * [Do users understand this hybrid checkbox/radio control?][5]
 * [Adding a “none of the above” option to a list of checkboxes][4]
 * [Combining radio and checkboxes][6]

[1]: https://ux.stackexchange.com/questions/70531/using-mixed-radio-and-checkbox-buttons-is-there-any-efficiency-or-usability-ga/82587#82587
[2]: https://i.sstatic.net/c2tIW.png
[3]: https://blog.codinghorror.com/usability-vs-learnability/
[4]: https://ux.stackexchange.com/q/67609/34594
[5]: https://ux.stackexchange.com/q/81525/34594
[6]: https://ux.stackexchange.com/q/102236/34594