The two examples you give are good ones because they represent two stages in the development of this trend. Both stem from the translation and transliteration of the Bible into Greek. This juncture in the transmission of non-Hebrew versions of the Bible has had many an impact in enduring transliterations that are strange to the modern English speaking ear.
In many cases, sounds that are impossible in the Greek alphabet were dropped or replaced by more Grecophilic ones. Take for example the dropping of the 'sh' (absent in Greek) from names like Simeon and Samuel, or the non-consonantal beginnings to names like Isaac and Israel (although that nuance is slightly more complex).*
In the present test case, it appears to us that no sounds were dropped and therefore the replacement of the simple 'f' sound with the more cumbersome /ph/ bigram is counterintuitive. It can be brought back to the intuitive realm by recalling some phacts about the sounds actually available to speakers of older Greek. The minimal pair of sounds spelled out with the letters /π/ (pi) and /φ/ (phi) was actually originally not the difference between 'p' and 'f', but between an aspirated 'p' and an unaspirated 'p'. These two letters were the obvious choices for differentiating between the two versions of the Hebrew letter /פ/ (pei). Then, when English came along and transliterated from the Greek, the aspirated version was spelled out, quite logically with an /h/ denoting the breath of air that accompanies the 'p' sound.
Thus, in a word that is almost exclusively Biblical, such as 'seraph', the proportion of /ph/ spellings to /f/ spellings is much more striking because most instances of it are [more] influenced directly by its appearance in the Bible. In contrast a word like 'aleph', whose transliteration is more likely to be direct (since most occurrences of it on the Internet today probably refer to the letter, a meaning not appearing in the Bible) would easily be spelled out the way they sound, with the letter /f/.
*Names are often the best exemplars of changes like this because they force transliteration as opposed to translation most of the time.
The assumption is that "originally" each consonant represented only one sound. Indeed, it makes little sense to devise an alphabet were a number of characters are intended to represent two separate consonants. There is some evidence for this, but putting that aside, it is interesting that even if this were so, already in biblical times at least one character indeed represented two sounds, the shin/ sin, which is clearly attested to in the shibboleth story in Judges.
As for p/ ph, it is a good transliteration convention. Although "f" and the sound /ph" are the same in English, only with /ph/ is the relationship with "p" apparent. Similarly, this is why it is good sense to use /kh/ for undageshed kaph, and even /bh/ for bet without the dagesh.
I'm not sure that Aleph or Seraph are good examples of Hebrew words "transliterated" into English, rather they are words which migrated into English from Hebrew. While these words may be pronounced more or less similar to the Hebrew counterpart they are more along the lines of words such as Josph. Such words are going to reflect nuances of the intermediary languages which they have passed through.
Transliterated words, those not at all part of the English language (or only recognized by select sub-groups such as those who frequent websites dedicated to the English language discussion of Judaic topics :) ) are probably much less likely to employ ph over f than words which migrated into English. Nevertheless when ph is employed it may often be due to the influence of migrated words (Joseph makes Yoseph more useful than it would be otherwise) directly or indirectly.
While I don't think there is an easy answer to why people choose ph of f (when they do) since we cannot read each individuals mind or account for style or other less concrete factors, Ph does have a small advantage over F insofar as it somewhat preserves the shared identity of peh and pheh in Hebrew (פ). This can be particularly useful when trying to illustrate the common root of two words where the pronunciation of the פ has shifted.