That's an interesting question. But I think any definitive answer will just be ignoring context and some of the dimensions involved.
Here's a breakdown:
Market survey
You'll find Hebrew and Arabic calendars that use either paradigm. Can't tell at which ratio, but a quick online search (especially if you search for the english translation) will reveal both.
Is it a Gregorian or Hebrew calendar?
The examples you have provided are Gregorian calendars, with Hebrew titles - the month starts on the 1st and ends on the 31st. This suggests users will already be familiar with the Gregorian calendar.
A Hebrew calendar will revolve around the Hebrew months and not the Gregorian ones. This will be (as one would expect) RTL:
(This wasn't easy to find, yet it has all the Gregorian components.)
Who are the users?
If we take Israel as an example, the majority of people have no idea what Hebrew day or month it is, and may only know the year because it is advertised around the Jewish new year. Most people won't be able to tell you their Hebrew birthday. Nearly every date communicated is Gregorian.
The exception to this is the orthodox:
Which are (surprise?) only 10% of the population and still most have mobile phones so they are exposed to the Gregorian calendar (to some extent).
Which leads us to the next point.
Pure RTLs are not common.
People who's main language is RTL, are still exposed to LTR content.
Probably first and foremost are numbers (which are LTR in both Hebrew and Arabic, even when Arabic numerals are used). Maths - as in equations, or how you'll be taught and develop solutions in exams - are also LTR.
Many of these people are exposed to other LTR languages (English, French in north Africa, etc.).
I'll be VERY surprised if in financial meetings - let it be in Egypt, Israel, or Lebanon - the spreadsheets are RTL and the time axis in graphs follows the same direction. But exceptions do exist:
So despite being RTLs, LTR orientation is something most poses.
Cognition - list inversions
Regardless of LTR orientation, it is worth noting that when it comes to direction inversions of lists - we don't find it too tricky.
Here are the first 10 prime numbers, RTL:
29 23 19 17 13 11 7 5 3 2
Can you tell which is the fifth prime number? Would it be any easier had the list been LTR?
This is in contrast to...
... sdrow ni srettel fo noisrevni ...
... where the brain loses the start-end/predictive shortcuts and resorts to reading each letter (at least a non-trained brain - there has been some research on this).
The argument here is, that if people have to read a list in inverted order, you may just as well make them do this on the day-of-the-week (of which there are 7), than on the day-of-month (of which there are approximately 30).
Use cases
Another important question is how people use this visual representation of a month.
Again, a key question is whether the Gregorian calendar is the basis for (non-visual) communication.
I'd say there are two primarily use case:
- Your appointment will be on Friday the 11th of October. In which case people would normally look for the 11th first - the search is done on the day-of-month. Same with Her birthday is on the 5th February.
- Dr. Moore only sees patients on a Thursday, which Thursday is good for you?. In which case people will start searching the days raw.
Which one do you think is more common?
There are obviously more use cases - I'll let the reader figure these out.
Summary
I think that if you read again through each section, given that your target audience is just generic RTLs, it seems more reasonable to have the day-of-month LTR (as they would be on an English Gregorian calendar), and have the days-of-the-week 'inverted'.
Trivia
Did languages switch to LTR in order to reduce smudging of ink? Answer.