Part 1 <cough, cough>
Yeah. I know. I am officially the G2G Word-Count Scofflaw. Sometimes I just can't help it. I'm all about sharing information. To ignore me, the PgDn button is your friend. :-)
I think the challenge you face, Alan, is that the devil is in the details. ;-) Unfortunately, AncestryDNA simply doesn't provide enough detail to make any (well, very many) but process-of-elimination judgments. All we get are their assumptions about cousinship distance plus the total cMs shared across however many segments. We can't plug that into DNA Painter or a spreadsheet or anything else to start analyzing where those matching segments likely came from.
I realize I'm gonna be doing some DNA 101 stuff in this post, so I don't want you to think I'm being patronizing; I don't know how much you've dabbled in this stuff yet and I don't want to skip something that may be useful.
When dealing with a combination of yDNA and autosomal DNA results, the atDNA analysis typically needs to come first. This is also the pathway for women to be able to use yDNA in their genealogical research: they can't contribute yDNA to test, so they need to establish the evidence linking them in proper association to a male who has taken a yDNA test.
Let's start with who--as I read things--is your best candidate: "ManB." If I'm correct, he's a Y-STR match with you at a genetic distance of 2 at 37 markers. You do share a common male ancestor. But at this point you can't know if that was your great-grandfather or a man born 12 generations ago. Unlikely, though, to be more than 10 generations. The Y-chromosome is really good at accurately determining patrilineal ancestry, and in diving back many generations with that accuracy. Not so good up close and personal. For example, I have a known 3C I'm one GD from, and a known 4C1R with whom I'm an exact match at 37 markers. However...
It sounds like both of you tested at FTDNA, yDNA and autosomal DNA. If he is also being displayed as an autosomal DNA match, the first things I would do are, A) get the detail from that atDNA match in FTDNA, and B) try to contact ManB for info about his genealogy. I'm sure you know this, but to get the match detail, from your profile go to his listing as a match. Click on the checkbox to the left of his name, then click on "Chromosome Browser." You get a nice graphic display...attractive but not much use by itself. If you hover your mouse over the segments where you match (usually shown on most browsers as a gold or orange color) you can see the details. Hard to copy them down from there, so click on "View this data in a table." You actually get more detail than you want. I personally would ignore anything below 10cM, but you can copy everything in the table and paste it elsewhere, or click on "Download to Excel" to save it as a .CSV file.
Now, there are no guarantees that the matching segments you see come only from your paternal line, even though you know you and he share a male ancestor. In fact, you can't even known that the matching segments come from your paternal line at all. That's one thing that makes autosomal DNA trickier to deal with than some folks realize. That said, what you do have is a reasonable hypothesis that those segments do come from your paternal line, so now you have to go about testing it.
Some people want "sciencey" DNA to give them quick answers about the unknown, but the truth is ya gotta put in the hard yards to identify the known first: baselines and benchmarks need to be established. There are a lot of nifty tools to help with this now. One I mentioned already is DNA Painter; another I recently bought but have only had minimal time playing with is called Charting Companion, in particular its "DNA Matrix" feature. Or you can just track everything in a spreadsheet.
The deal is that, as you've shown you know by your question, one data point does not a series make. And yDNA and autosomal DNA are apples and oranges...until you have the data confirmation that links them. First step is to create a database (even a free Google sheet works fine) with your known matches, starting from the closest--excluding parents and full siblings: too much relatedness there. If you have a good paper-trail on the individuals, then they're "known." In there needs to be detailed segment information, so--unfortunately as always--AncestryDNA data are out of the picture. Total shared centiMorgans isn't enough; you have to know the chromosome numbers, the segment sizes in cMs, and the positions as starting and ending base-pair numbers. The "known" part means that you should be able to understand where the matching segments to you are likely to have come from. Layer 1, you separate your father's and your mother's lines; layer 2 and you can start determining which set of the grandparents; layer 3 and you've got a framework for which specific grandparent because you've now determined which set of g-grandparents, and so on.
An aside here. If using GEDmatch, which I do recommend, I suggest starting off running their "Are my parents related?" tool. Sounds odd, and maybe even a little insulting. But what it does is look at what's called ROH, runs of homozygosity: contiguous stretches of chromosomes where the two nucleotide values are the same, e.g., A-A, C-C. This is an indication of pedigree collapse; the greater the amount of ROH, the more recent the "collapse." Every family tree has it somewhere--actually in multiple places--but how prevalent it is in your own data becomes a factor in how you evaluate and weigh the size of shared segments.